Fasted and fed state human duodenal fluids: Characterization, drug solubility, and comparison to simulated fluids and with human bioavailability

Accurate in vivo predictions of intestinal absorption of low solubility drugs require knowing their solubility in physiologically relevant dissolution media. Aspirated human intestinal fluids (HIF) are the gold standard, followed by simulated intestinal HIF in the fasted and fed state (FaSSIF/FeSSIF). However, current HIF characterization data vary, and there is also some controversy regarding the accuracy of FaSSIF and FeSSIF for predicting drug solubility in HIF. This study aimed at characterizing fasted and fed state duodenal HIF from 16 human volunteers with respect to pH, buffer capacity, osmolarity, surface tension, as well as protein, phospholipid, and bile salt content. The fasted and fed state HIF samples were further used to investigate the equilibrium solubility of 17 representative low-solubility small-molecule drugs, six of which were confidential industry compounds and 11 were known and characterized regarding chemical diversity. These solubility values were then compared to reported solubility values in fasted and fed state HIF, FaSSIF and FeSSIF, as well as with their human bioavailability for both states. The HIF compositions corresponded well to previously reported values and current FaSSIF and FeSSIF compositions. The drug solubility values in HIF (both fasted and fed states) were also well in line with reported solubility data for HIF, as well as simulated FaSSIF and FeSSIF. This indicates that the in vivo conditions in the proximal small intestine are well represented by simulated intestinal fluids in both composition and drug equilibrium solubility. However, increased drug solubility in the fed vs. fasted states in HIF did not correlate with the human bioavailability changes of the same drugs following oral administration in either state.


Introduction
A systemically acting drug administered orally must be absorbed from the gastrointestinal (GI) tract and avoid first-pass extraction in the gut wall and liver before it can exert its pharmacological effect.The fraction absorbed from the intestine is determined by the velocity (length/time) of drug transport across the apical intestinal cell membrane, i.e. the permeability.It is also determined by the drug dissolution rate and solubility in the intestinal lumen, because only aqueous dissolved drug molecules cross the intestinal epithelial barrier.
Thus, the rate-limiting step in intestinal drug absorption can be either the solubility/dissolution rate of the drug in the GI luminal fluids or the intestinal permeability.In drug development, solubility is the more frequently observed limitation, because candidate drugs in drug discovery and early development tend to have limited aqueous solubility and slow dissolution rate in the GI lumen [1].This is related to the lead-finding techniques applied, which often are based on molecular in silico and in vitro screening of receptor-ligand interactions [2][3][4].These screening tools select for drugs that can interact with target receptors, meaning they will typically be hydrophobic and/or lipophilic [5][6][7].
Consequently, low aqueous solubility leading to low oral bioavailability is a bottle neck in drug discovery and early development even though this biopharmaceutical factor was recognized several decades ago.Drug solubility and dissolution rate in water are therefore critical biopharmaceutical parameters for any candidate drug or drug product and are investigated by a range of methods [8].Early development requires resource-and time-efficient methods to identify drugs with unfavorable biopharmaceutical and pharmacokinetic properties, such as poor intestinal solubility.This typically happens when the solubility value in water is below a predetermined value, for example, a drug with a dose number (Do) > 5, as proposed by Lipinski [9].However, highthroughput water solubility values of a drug do not necessarily reflect the compound solubility in the GI lumen.Especially for lipophilic drugs, the difference can be substantial.For instance, cinnarizine, fenofibrate, and danazol are three low-solubility drugs (<5 µg/mL in water), but they have 20-fold higher solubility in biorelevant fasted state intestinal media than in plain aqueous buffer at pH 6.5 [10].
Hence, it is important to investigate drug solubility in media that give accurate in vivo absorption predictions [11].The gold standard in biopharmaceutical drug solubility assessment is human intestinal fluid (HIF).By individual sampling of different parts of the GI tract at different prandial states, HIF takes into account the large intra-and inter-individual variabilities.This is important because food intake triggers neural and hormonal signaling from the GI tract in response to gastric distension and the chemical presence of nutrients.For instance, digested dietary lipids in the lumen control GI motility, luminal pH, bile salts, and lipase secretion [12,13].The fasted state HIF composition is also affected by the difference in gastric, intestinal bile and pancreatic secretions during the human interdigestive motility cycle [14,15].However, HIF is not practical for routine solubility investigations in preclinical drug development as it involves sampling from human subjects, a time-and resource-demanding approach.An alternative to HIF is simulated intestinal fluid (SIF) in the fasted (FaSSIF) and fed (FeSSIF) states.These in vitro compositions of bile salts and phospholipids at appropriate pH, buffer capacity and osmolarity replicate the luminal GI conditions to better capture in vivo drug solubility and drug product performance.During the last two decades SIFs have been refined and updated to identify which luminal components primarily affect the solubility of different drugs [16,17].Nonetheless, there is still controversy regarding their accuracy in predicting HIF drug solubility.This applies especially to low-solubility compounds (<10 µM) in FaSSIF, and for FeSSIF, where current data indicate an underestimation of HIF solubility values [18].In part, this is due to the limited number of drugs that have had their solubility determined in HIF and to the version of SIF used in the solubility correlations.Further, data on HIF characterizations are variable because of lack of consensus on the in vivo components and their concentrations, different food, drink, and sampling protocols, and the complexity of neural and endocrine responses in the GI tract [19].
This investigation aimed at characterizing pooled fasted and fed state HIF from the duodenum of 16 volunteers with respect to pH, buffer capacity, osmolarity, surface tension, total protein and phospholipid content, and bile salt compositions and concentrations.The fasted and fed state HIF samples were then used to investigate the equilibrium solubility of 17 representative low-solubility small-molecule drugs, whereof six were confidential industry compounds and 11 were wellknown model drugs.These 11 were chosen for their structural diversity after a comprehensive analysis.Next, the solubility values from this study were compared to reported solubility values in fasted and fed state HIF, FaSSIF, and FeSSIF, as well as to their human bioavailability when administered in the fasted and fed states.This was to evaluate the reproducibility of HIF solubility values, the accuracy of SIFs for predicting solubility in HIF, and the potential of using HIF drug solubility for predicting systemic exposure after oral dosing.

Molecular diversity analysis
A dataset of 691 pharmaceutical compounds used on the Swedish market in the year 2001 was used as reference for the molecular diversity analysis.Preparation and description of this dataset has been published elsewhere [20,21].Permanently charged molecules (n = 17) in the reference dataset were not included in the analysis herein.Structures were imported in Maestro (Schrödinger Release 2019-3: Maestro, Schrödinger, LLC, New York, NY, 2019).The dataset of 11 model compounds was compiled from the respective structures if they were present in the reference dataset.If not present, they were imported separately in Maestro as SMILES.The structures were prepared using LigPrep (Schrödinger Release 2019-3: LigPrep) in their neutral form.Descriptors were calculated using QikProp (Schrödinger Release 2019-3: QikProp) in default mode.Principal component analysis was performed using default settings and visualized in SIMCA (version 15, Umetrics, Umeå, Sweden).Included QikProp descriptors in the principal component analysis were (descriptor abbreviation found in the PCA loading plot are given in parentheses): molecular weight (mol MW), number of hydrogen bond acceptors (acceptHB), number of hydrogen bond donors (donorHB), solvent accessible volume (volume), solvent accessible surface area (SASA), hydrophilic component of SASA (FISA), hydrophobic component of SASA (FOSA), weakly polar component of SASA (WPSA), carbon atom π-system component of SASA (PISA), Van der Waals surface of polar nitrogen and oxygen atoms (PSA), globularity (glob), and predicted octanol/water log P (QPlogPo/w).

Subjects and intestinal sampling procedure 2.3.1. Study subjects
The study was approved by the regional ethics committee for human research in Uppsala, Sweden (no: 2013/029) and Leuven, Belgium (S53791).Sixteen healthy male (n = 10) and non-pregnant female (n = 6) volunteers, 18-45 years of age who had not been exposed to frequent ionizing radiation in the previous year were eligible to participate.Before signing informed consent and enrolling in the study, volunteers were tested against inclusion and exclusion criteria.A clinical physician assessed volunteers to be healthy based on a physical examination (free from illness or gastrointestinal disorders) and laboratory tests (hepatitis B, hepatitis C or HIV).This study was conducted at Uppsala University Hospital, Sweden, and Leuven University Hospitals, Belgium, in accordance with the International Conference of Harmonization guidelines for good clinical practice and the principles described in the World Medical Association Declaration of Helsinki.

Study design and sampling procedure
This intestinal sampling study was a crossover trial performed in the fasted and fed states in 16 subjects: eight each at Uppsala University and University of Leuven.After an overnight fasting period of at least 10 h, all volunteers arrived at the hospital.Following topical anesthesia of the nose-throat cavity with Xylocaine (2% gel or 10% spray), a doublelumen polyvinyl catheter fitted with a guide wire (Salem Sump Tube 14 Ch, external diameter 4.7 mm) was inserted through the nose and placed in the duodenum (third or fourth part).The GI position was checked by fluoroscopy and verified by pH and visual inspection of a small volume of intraluminal fluid.The intubation was followed by a 20min stabilization period, after which fasted subjects drank 240 mL of tap water (t = 0).The samplings started with the fasted state, and the fed state sampling was performed on the same day directly after the fasted state sampling at Uppsala University, and at least one day after at the University of Leuven.
The fasted state sampling started 10 min after intake of water, and intestinal samples were collected by suction through the tube at 10-min intervals during 120 min using a syringe.The fed sampling started the subject drinking 400 mL Ensure Plus nutritional supplement drink (600 kcal, 29% lipids, 54% carbohydrates, 17% proteins) within 10 min.At 20 min after the start of meal intake, fed subjects drank 240 mL of tap water (t = 0).
The fed state sampling procedure followed the same protocol as the fasted one described above.After the last sampling time (t = 120 min) the GI catheter was removed.All duodenal samples were immediately weighed, analyzed for pH, and stored on ice.To prevent lipolysis, the fed state duodenal fluids were immediately added the lipase inhibitor (Orlistat 1 μM).Samples were centrifuged (4 • C, 10 min, 4000 × g) to remove debris, whereupon all samples from each individual and prandial state were pooled (2 pooled samples each per individual, fasted and fed state).The pooled individual samples were stored at − 20 • C and shipped to Sanofi R&D (Vitry, France) for detailed characterization and further pooling (2 population pools, fasted and fed state).Pooled samples were characterized with respect to: collected volume, pH, osmolarity, buffer capacity, surface tension, total protein, phospholipid and bile salts content, and bile salts composition, as described in detail below.

Human intestinal fluids physicochemical characterization
HIF were characterized in physicochemical terms of pH, buffer capacity, surface tension, and osmolarity.The first two of these were measured at 37 • C and the latter two at ambient temperature.pH was measured using a pHM240 meter (Radiometer, Copenhagen, Denmark) and the buffer capacity was determined by titration on the same equipment.The surface tension was determined with the Wilhelmy plate method (Thermo Cahn 322 Dynamic Contact Angle Analyzer, Irvine, CA, USA).The osmolarity was determined on a Vapro vapour pressure osmometer (model 552O, Wescor Inc., Logan, UT, USA).

Drug solubility in human intestinal fluid aspirates
The equilibrium drug solubility in intestinal aspirates was performed on 17 small-molecule compounds covering a broad chemical space.Eleven were well-known model drugs and six were confidential industry compounds.Solubility experiments were performed in duplicates (n = 2) in both fasted-and fed-state pooled HIF for each compound.
The solubility study and the analytical characterization of samples were carried out in Biosafety Level-2 laboratories.The pooled HIF samples were stored at minus 20 • C. Before the solubility determinations, small volumes of intestinal fluids were separately frozen to decrease the number of thawing processes of the main pooled intestinal fluid containers.After a 60 min thawing process, the fasted and fed HIF were weighted into vials, with 1 mL in each.To these vials an excess API amount of 2 mg was added to obtain a saturated suspension with the target concentration of 2 mg/mL.All API containing vials were put on a shaker for agitation for the duration of 24 h at 37.0 ± 0.5 • C under light protection.At each of the seven sampling times (15, 30 and 60 min, and 2, 3, 6 and 24 h) the shaking was stopped and 50 µL samples were taken and filtered using a microplate filtration process before the sample analysis.After the sampling process the sampled volume was replaced with HIF, and agitation was continued to reach the next sampling point.

D. Dahlgren et al.
The study was stopped after 24 h where also the pH of the samples were measured.A comparative study was performed to evaluate the solubility determinations after filtration (PTFE, 0.45 µm pore diameter) and centrifugation processes, but there were no significant differences.The equilibrium solubility was defined as when the dissolution reached a steady-state plateau (±5% between time points) or at 24 h.

Solid state characterization
The X-ray powder diffraction pattern of the solubility residues was compared with the pattern of the relevant solid phase API.This analysis showed no changes in solid state for any of the APIs, exemplified for Felodipine in Supplementary file 1. Analyses were carried out at room temperature on a Brucker D4 Endeavor instrument using the Bragg-Brentano (vertical θ-θ configuration) parafocusing geometry.A sealed copper anode X-ray tube running at 40 kV and 35 mA was used (λ CuKα average = 1.54178Å).A counting time of 1 s per step in an angular range [2 • − 40 • ] with a 0.016 • step size in 2θ was used for each shortterm sample analysis.For each experiment, the powder was deposited onto the surface of a sample holder.Then the sample was sealed with a Kapton® film for safety consideration.

Comparisons
Drug solubilities in the fasted and fed state HIF from this study were compared to those from the literature, and to FaSSIF and FeSSIF.As the reported SIF solubility values were determined in different FaSSIF and FeSSIF versions, their compositions are presented in Table 2. Solubilities in fasted and fed state HIF were also compared to the drug bioavailability or plasma exposure in the two states.

Bioanalyses
Protein concentration was measured using a Roche Cobas Mira clinical chemistry analyzer (Basle, Switzerland) and a commercial enzyme-based colorimetric assay (Bio-Rad Protein Assay; Bio-Rad Laboratories, Hercules, CA, USA).The analytical method used for the quantification of bile salts and phospholipid content in HIF is described in detail by Riethorst et al, 2016 [22].Drug HIF concentrations were analyzed with a validated analytical method at Sanofi, France, on an Agilent 1290 Infinity 1 type HPLC equipment with a ZORBAX Eclipse Plus C18 (2.1*50 mm 1.8 µL) column: H 2 O + HCOOH 0.05% (A) / ACN + HCOOH 0.035% (B), 1.1 mL min, 55 • C, runtime 2.2 min (1.5 min 98% A, 0.5 min 98% B, 0.2 min 98% A).The standard calibration curve had a linearity of R > 0.999 for all compounds.

Statistical analysis and calculations
All results are presented as individual values or mean ± standard deviation (SD) or as the ratio between the solubility in the fasted and fed states (ratio = average fed/average fasted) for both HIF and SIF.Cut-off values for a difference in fed vs. fasted solubility was set to 3 < ratio < 0.33, based on the anticipated variability in solubility determinations between laboratories.This inter-laboratory variability depends, in turn, on the experimental protocol (e.g.amount of drug powder used) and lack of standardization of data analysis [23].When there was more than one reported solubility value in HIF or SIF, the lower value was used in the correlations.Do was calculated using equation (1), by relating the maximum oral dose and gastric volume (250 mL) to the drug solubility in fasted and fed HIF.
This Do was theoretical and not related to the regulatory-defined Do based on aqueous solubility in the pH range of 1.0-6.8.A Do > 1 predicts that the intestinal volume is insufficient to dissolve the maximum dose, whereas Do < 1 predicts that all drug will be in solution in the intestines.

Molecular diversity of the drug data set
Principal component analysis of the reference dataset with 674 compounds and 12 physicochemical descriptors resulted in two principal components.In total, this explained 73% of the variance in the dataset.The score-and loading plots are presented in Fig. 1a and b, respectively, where visual analysis of the loading plot suggests that the general property directions in the plots are molecular size in principal component 1 (x-axis) and hydrophobicity in principal component 2 (yaxis).The predicted score plots of the 11 known compounds with HIF solubility in the fasted and fed states are presented in Fig. 1c and d, respectively.

Human intestinal fluids characterization and composition
Characterizations of fasted-and fed-state HIF (described below) were performed on pooled HIF samples from Uppsala University and University of Leuven.The mean volumes (±SD) of HIF collected in each individual during 120 min were 45.8 ± 14.3 and 59.2 ± 28.3 mL in the fasted and fed states, respectively, and there were only minor differences between the two sites.

Duodenal pH, buffer capacity, osmolarity, surface tension, and total protein, phospholipid, and bile salt contents
Table 3 presents pH, buffer capacity, osmolarity, surface tension, and total protein, phospholipid, and bile salt content of the pooled fastedand fed-state HIF samples.The fed-state duodenal pH was slightly lower than the fasted-state pH (5.96 vs. 6.83), and two parallel pH stability investigations one month apart on frozen samples showed no change in pH over 8 h.There was only a minor difference in surface tension of the fasted (34.3 mN/m) and fed (26.4 mN/m) samples at 37 • C. The fedstate values were substantially higher than the fasted ones for buffer capacity at 37 • C (3.3 fold), osmolarity (2.0 fold), and total protein (5.9 fold), total phospholipid (19.5 fold), and total bile salt (2.5 fold).

Duodenal bile salt composition
The bile salt concentrations and composition in the fasted and fed states are presented in Fig. 2.There were only minor differences in the qualitative composition between the two prandial states.The most pronounced difference (fraction vs. fraction) was for taurocholic acid (5.2%), followed by glycochenodeoxycholic acid (2.5%), glycodeoxycholic acid (2.7%), and taurochenodeoxycholic acid, taurodeoxycholic acid, glycoursodeoxycholic acid (~1% each), while there were no differences (<0.5%) for glycocholic acid and tauroursodeoxycholic acid.

Table 3
The mean (±SD) pH, buffer capacity, osmolarity, surface tension, and total protein, phospholipid, and bile salt content of the pooled (n = 16 individuals) fasted and fed state human intestinal fluid (HIF) samples.

Equilibrium solubility
The two individual equilibrium solubility (n = 2) measurements of the model drugs in pooled fasted-and fed-state HIF are presented in Table 4.The variability (CV%) in the drug solubility values obtained from the two determinations were below 30%, with the exception of the least soluble drug, bromocriptine, in the fasted state (46%).For reference, Table 4 also includes the reported mean (±SD), or range, solubility values of the same drugs in fasted-and fed-state HIF.
Fig. 3A shows the ratio between drug solubility in fed-vs.fasted-state HIF for the 17 drugs.For seven of them (A1150, A7651, ibuprofen, valsartan, A4356, A8942, and tadalafil) there were no difference between their fed vs. fasted solubility ratios (defined as a fold difference in the range of 0.33 to 3).The remaining ten drugs had a higher solubility in the fed state (ratio > 3), with the following rank order (low to high): A9530, zafirlukast, carvedilol, felodipine, probucol, aprepitant, ketoconazole, bromocriptine, A1260, fenofibrate.Of these ten drugs, eight had a substantially higher fed/fasted solubility ratio (>10).Drugs with a lower fasted-state HIF solubility tended to have increased solubility in the fed state (Fig. 3B).This was especially apparent for drugs with a solubility in fasted HIF below 30 µg/mL.

Comparison with reported drug solubility values in human intestinal fluids
Table 4 shows the reported equilibrium solubility in fasted-and fedstate HIF for 7 and 4 drugs, respectively, that were also included in the present study.Reported values are compared to the ones obtained in this study in Fig. 4A.There were only minor differences between the reported solubility values and the study ones, for fasted and fed states.The only drug for which there was a difference (>3 fold) was fenofibrate in the fasted state.

Comparison with reported drug solubility values in FaSSIF and FeSSIF
Table 5 shows reported equilibrium solubility values in FaSSIF (11 drugs) and FeSSIF (10 drugs, no value for probucol).For comparison, the fasted-and fed-state mean (±SD) equilibrium solubility of the model drugs in HIF, and the reported FaSSIF and FeSSIF solubility values, are presented in Fig. 5A and B, respectively.
In the fasted state, only valsartan out of the 11 drugs had a higher solubility (>3 fold) in fasted HIF than FaSSIF, while 5 drugs (bromocriptine, zafirlukast, fenofibrate, felodipine, carvedilol) had a lower solubility (<0.33 fold) in fasted HIF than in FaSSIF.In the fed state, 3 out of 10 drugs (fenofibrate, ketoconazole, valsartan) had a higher solubility (>3 fold) in fed HIF than FeSSIF, and only bromocriptine had a lower solubility (<0.33 fold) in fed HIF than FeSSIF.In total, there were two GCDC 879 µM 25

Table 4
The two (A and B) individual equilibrium solubility measurements of 17 model drugs in pooled fasted and fed state human intestinal fluids (HIF).Presented are also previously reported mean ± SD (or range) solubility values of some of the drug compounds in fasted and fed state HIF [18,51].drugs (bromocriptine and valsartan) for which the values deviated>10 fold between HIF and SIF; regardless of prandial state, valsartan had higher solubility in HIF, and bromocriptine had a lower solubility in HIF than in SIF.Fed-vs.fasted-state solubility ratios in HIF and SIF are compared in Fig. 6.The solubility ratio in HIF was higher (>3 fold) than in SIF for 3 out of 10 drugs (zafirlukast, bromocriptine, fenofibrate), i.e., SIF underpredicted the food-induced increase in HIF solubility.This underprediction was primarily related to a higher solubility in FaSSIF than in fasted HIF (Fig. 5A, bromocriptine, zafirlukast, fenofibrate, felodipine, carvedilol), and to a lesser extent a higher solubility in fed HIF than in FeSSIF (Fig. 5B, only bromocriptine).

Dose number in fasted and fed state HIF
Table 6 shows the Do for 11 model drugs.Five of the drugs (aprepitant, carvedilol, ketoconazole, bromocriptine and fenofibrate) had their Do reduced from values above 10 in fasted HIF to below 5 in fed HIF.A high permeability drug with a Do below 5 is defined as a lowsolubility compound, as proposed by Lipinski [9].

Discussion
This study collected and characterized duodenal HIF from 16 healthy volunteers in fasted and fed states with respect to pH, buffer capacity, osmolarity, surface tension, total protein and phospholipid content, and bile salt compositions and total content.Pooled HIF samples were used to determine equilibrium solubility of 17 low-solubility model drugs (11 known and 6 confidential) in fasted and fed states.The 11 known drugs were physicochemically characterized, and their solubility values in HIF were compared to reported values of: i) HIF solubility, ii) human bioavailability in the fasted and fed states, and iii) FaSSIF and FeSSIF solubility.
Principal component analysis of the 11 known drugs predicted them into the chemical space of 674 marketed drugs [20,21].The 11 drugs were fairly well distributed in the hydrophobic chemical space without any extreme data points compared to the marketed drugs; see the score plots in Fig. 1.Although there are directions in the score plot of the fasted-state HIF solubility that hold the relatively most soluble (ibuprofen and valsartan) and the least soluble compounds (bromocriptine and zafirlukast), no general trend for HIF solubility was evident  4).(B) Comparison between the increases in fed HIF vs. fasted HIF ratio compared to the solubilities of the drugs in fasted HIF.
D. Dahlgren et al. for the investigated physicochemical properties.
Luminal pH (and buffer capacity) has an impact on the solubility of weak acidic and basic drugs.The duodenal pH values in this study were 6.83 and 5.96 in the fasted and fed states, respectively.This is well in line with human luminal pH values from three other aspiration studies in the proximal small intestine of healthy volunteers, which all report a slightly higher pH in the fed than in the fasted state (e.g.7.5 vs. 6.1 [34], 6.78 vs. 6.22 [22], 6.3 vs. 6.1 [35]).An exact proximal small intestinal pH value is not readily defined, as the value is affected by the position of the intestinal tube (pH increases aborally from the pyloric sphincter [35]); the interdigestive motility complex in the fasted state [14]; and the dietary composition in the fed state [36].Of these three factors, the luminal position is probably more important than the pH and buffer capacity of the food.This was supported by the higher pH observed in the distal duodenum compared to the proximal jejunum (6.0 vs. 7.4, although only a few centimeters apart) after administration of comparable nutritional drinks [34,37] Comparison of fasted and fed state HIF to FaSSIF and FeSSIF, respectively, showed a high similarity between pH and buffer capacity, except for FeSSIF-V1, for which the pH was lower (5.0 vs. 5.96) and the buffer capacity higher (76 vs. 18 mmol/L/pH).
Osmolarity, like pH, is a luminal parameter under strict physiological regulation in vivo.The intestine rapidly reacts to variations in osmolarity to keep the intestinal fluid isotonic.Depending on luminal tonicity, this is done by either secretion or absorption of osmolytes, coupled to passive water flux [38].For instance, the luminal osmolarity increased from 50 to 170 mOsm in the first 25 cm of the proximal small intestine following intake of water, a process that occurs during approximately 3 min [39,40].A comparison of the fasted and fed state osmolarity values in this study (189 and 372 mOsm) to those in FaSSIF and FeSSIF, showed that our values closely resemble those in version 2 (180 and 390 mOsm).Given the rapid luminal adjustment of osmolarity and its overall minor impact on the solubility of small drug molecules [16], it is probably not so important in drug solubility investigations.
In this study, the surface tension of fasted-and fed-state HIF were comparable (34.3 and 26.4 mN/m) and similar to reported fasted state values (≈28 mN/m) [34].All versions of FaSSIF (≥54 mN/m) and FeSSIF (≥40.5 mN/m) consequently overpredicted surface tension by at least 15 mN/m [41].This overprediction may result in an increased drug dissolution rate of lipophilic compounds, as surface tension reflects the wetting behavior, whereas the effect of surface tension on equilibrium drug solubility should be minor [42].Still, luminal amphiphilic components (e.g., bile salts, fatty acids and phospholipids) in HIF may still affect solubilization, and thereby increase the equilibrium solubility of the same drugs (an effect unrelated to particle wetting).Luminal protein content is low in the fasted state and in the fed state primarily reflects meal composition.In this study, the duodenal protein concentration increased from 2.8 ± 0.4 to 16.5 ± 0.2 g/L with 400 mL Ensure Plus (protein content 17 mg/mL).This can be compared to a rise in the jejunum from 1.0 ± 0.1 to 5.0 ± 0.1 g/L following 200 mL of Nutriflex (protein content 19 mg/mL) [34].The difference in fed-state protein content between the duodenum and jejunum after ingestion of similar protein drinks illustrates the large capacity of the upper GI tract to digest and rapidly absorb protein.The rapid luminal absorption of digested proteins and its insignificant impact on drug solubility may explain why proteins are not included in any of the FaSSIF or FeSSIF versions.
Bile salts and phospholipids (with or without fatty acids) form mixed micelles (6.5 nm in diameter) and vesicles (50 nm in diameter) in the intestinal lumen, where the micelles predominate at high concentrations (i.e., fed state) [43].Both micelles and vesicles facilitate the solubilization of lipophilic vitamins and drugs, and are thus important for in vivo relevant drug solubility determinations.Reported bile salt concentrations in the fasted state range from 1.4 to 8.1 mM, and from 3.6 to 24.0 mM in the fed state [22].The fasted-and fed-state values from this study (3.52 and 8.91 mM) fell in the middle of the reported luminal concentration ranges.Total bile salts concentration was thus well captured in all versions of FaSSIF and FeSSIF, but with a better similarity in FeSSIF-V2 than FeSSIF-V1.While the total bile salts content changed with prandial state, the composition was essentially unaffected.These results thus corroborate data from others who report similar bile salt compositions in any one individual over time [22].The corresponding reported values for phospholipids in the fasted state range from 0.1 to 1.8 mM, and from 1.2 to 6.0 mM in the fed state [22].The fasted HIF phospholipid level in this study (0.19 mM) fell in the lower part of reported values and the fed value (3.72 mM) in between.For phospholipids, the concentrations in FaSSIF-V2 capture the fasted HIF value better from this study, whereas FeSSIF was better than FeSSIF-V2.In summary, given the individual variability and the small differences in concentrations for both bile salts and phospholipids between HIF and SIF, reassessment of these parameters in SIF need not be prioritized.
Duodenal fluids in this study were collected and pooled during 120 min following intake of 240 mL water (fasted state) or 400 mL of a nutritional drink followed by 240 mL of water (fed state).It should be mentioned that values can vary depending on individual variations in luminal composition and the time of sampling after intake [22].However, there was an overall good correspondence between the pooled values in this study and reported pooled or mean/median HIF values in the literature.There was also a good correspondence of HIF in the fasted and fed states to FaSSIF and FeSSIF, especially for the later version of the SIFs, which better captured the upper small intestinal pH, buffer capacity, osmolarity, and total bile salt content (Table 2).The luminal composition data from this study supported the current compositions for FaSSIF and FeSSIF.However, their texture and viscosity could still be improved, to better mimic intake of a solid meal rather than a liquid one.Luminal viscosity is a parameter that may impact the in vivo dissolution rate and/or drug product performance following oral intake [44].The next level of in vitro dissolution models should also better reflect the dynamic changes in the fasted state as well as during digestion.It would also be clinically important to have improved GI models for certain GI diseases [19].
For seven drugs in the fasted state and four in the fed state in this study, there were previously reported equilibrium solubility values in HIF (Fig. 4A).With one exception (fenofibrate in the fasted state), this comparison showed only minor differences, regardless of the different food and dosing regiments in the reported studies.This indicated that reproducibility was acceptable for drug equilibrium solubility determinations between laboratories using different aspiration techniques and types of meals [18].Further, the HIF solubility values in the fed state were for the same, or higher, than in the fasted state for all the drugs.The drugs tended to have increased insolubility in the fed state if their solubility was low in the fasted state.This is logical given the inverse correlation between drug lipophilicity and water solubility, where the addition of lipophilic constituents has a larger effect on the solubilization of lipophilic (i.e.low solubility) compounds.Based on the limited dataset in this study, the fed state had an insignificant impact on the equilibrium solubility of those drugs with a fasted state solubility above 30 µg/mL.
With the exception of the solubility values of valsartan and bromocriptine in fasted and fed state HIF, equilibrium solubility values in HIF and reported SIF were within a 10-fold difference.This was acceptable considering the large interlaboratory variability (CV% 〈1 6 0) of the water solubility values reported by Andersson et al., 2016, for the same compounds and identical experimental conditions [23].The large HIF vs. SIF difference for valsartan and bromocriptine cannot be readily explained by any compositional differences, or by any unique properties of these two drugs.It is likely that the unusually large differences in the reported SIF values are associated with the experimental method of drug solubility determination/quantification in those studies, as there were no difference between the HIF solubility values in our study and reported HIF solubility values for either the fasted or fed states.It was also evident that the different ratio of fed vs. fasted solubility was higher in HIF than in SIF, an effect generally related to a higher solubility in FaSSIF than in fasted HIF.This cannot be attributed to any dissimilarities in composition between these two fasted media, as they had only minor differences in pH, buffer capacity, and phospholipid and bile acid concentration.Based on the data in this study, it can be concluded that current SIF media are generally successful in predicting HIF solubility.However, care must be taken when interpreting FaSSIF solubility data, as these tended to slightly overpredict fasted-state HIF solubility, especially for low-solubility drugs (Do > 5, which applies to all drugs except ibuprofen and valsartan).These results are in opposite of previous observations, which indicate that FaSSIF slightly underpredict HIF solubility for low-solubility drugs [18].
Interestingly, the increase in fed-vs.fasted-HIF solubility in this study (accompanied by a decrease in Do) was not reflected in increased human bioavailability or exposure of the same drugs following dosing of immediate release formulations in fasted and fed states.The results are in contrast to the theory that an increase in solubility of a poorly soluble compound should also increase its absorption [45].The dissimilarity may be related to the free-dissolved drug concentration, or the molecularly dissolved one, in the water phase of HIF/SIF; these may be unaffected by addition of solubilizing constituents, such as bile acids, fatty acids and phospholipids.Further, the rapid absorption of fatty acids and phospholipids in the upper small intestine typically causes an immediate precipitation of liberated drug molecules thereby limiting the timewindow for supersaturation.Consequently, the data presented here suggest that an increased equilibrium solubility in fed HIF, compared to fasted HIF or water, should not be interpreted to be predictive of absorption of low solubility compounds.Other biopharmaceutical luminal processes must also be taken into account, such as the size-related behavior of drug crystals, precipitation/supersaturation, and the disintegration of various solid dosage forms.Likewise, the effect of prandial state on physiological parameters affecting drug absorption (such as gastric emptying rate and complexation with food and luminal components) must be included.

Conclusion
This study generated a broad physicochemical and compositional analysis of HIF in fasted and fed states and coupled these to equilibrium drug solubility measurements for a range of low-solubility drugs.The 11 known model drugs are well embedded in the lipophilic part of the physicochemical space of marketed drugs.The HIF compositions corresponded well to previously reported values and to current FaSSIF and FeSSIF compositions.The drug solubilities in HIF (both fasted and fed states) were also well in line with reported solubility values for HIF and simulated FaSSIF and FeSSIF.This indicates that the in vivo conditions in  4), as well as reported values in fasted (FaSSIF) and fed (FeSSIF) states of simulated intestinal fluids (Table 5).A (-) indicates a higher (3-fold) solubility in simulated fluids than in HIF, and (+) indicates a higher (3-fold) solubility in HIF than in simulated fluids.D. Dahlgren et al. the proximal small intestine are well reflected by SIF with regard to both composition and equilibrium solubility.There was, however, no correlation between HIF drug solubility ratios in the fed vs. fasted states and human bioavailability ratios of the same drugs in either state, following oral administration, indicating that for these drugs the solubility in the fed state is not alone determining the food effect.4) with reported ratios in fasted (FaSSIF) vs. fed (FeSSIF) state simulated intestinal fluids (Table 5).The solid line shows the ideal agreement.The upper and lower dashed lines show where the simulated intestinal fluids overpredics (>3 fold) and underpredicts (<0.33 fold) the increase in fed/fasted HIF solubility, respectively.

Table 6
Dose number (Do) of 11 of the known model drugs.Here Do represents the relationship between the maximum oral dose dissolved in 250 mL (representing a glass of water) and the mean equilibrium solubility in fasted and fed state human intestinal fluids (HIF) presented in Table 4 (i.e., maximum dose/250 mL/ HIF solubility).The compounds that change from a low solubility compound (Do > 5 [9]) in the fasted state to a high solubility one in the fed state are indicated by a star *.

Fig. 1 .
Fig. 1.Principal component analysis results.The score plot ellipse is the confidence region based on Hotelling's T 2 (95%) implemented in SIMCA as outlier indicator.(a) Score plot of the reference dataset.(b) Loading plot of the reference dataset.Predicted score plot of the 11 known compounds investigated, colored according to human intestinal molar solubility in (c) fasted state and (d) fed state.

Fig. 3 .
Fig. 3. (A)Ratio between the mean drug solubility in the fasted compared to fed (fed/fasted) state human intestinal fluids (HIF) as reported in this study (Table4).(B) Comparison between the increases in fed HIF vs. fasted HIF ratio compared to the solubilities of the drugs in fasted HIF.

f e n o f i b r a t e p r o b u c o l a p r e p i t a n t f e l o d i p i n e c a r v e d i l o l k e t o c o n a z o l e i b u p r o f e n f e n o f i b r a t e f e l o d i p i n e k e t o c o n a z ou p r o f e n v a l s a r t a n t a d a l a f i l z a f i r l u k a s t c a r v e d i l o l f e l o d i p i n e a p r e p i t a n t k e t o c o n a z o l e b r o m o c r i p t i n e f e n o f i b r aFig. 4 .
Fig. 4. (A) Comparison of the mean solubility in the fasted and fed state HIF in this study with reported values[18].A (-) indicates a higher (3-fold) solubility in reported HIF compared to this study.(B) Comparison between the solubility ratios of fed vs. fasted state HIF in this study with reported bioavailability values for these two states.

b r o m o c r i p t i n e z a f i r l u k a s t f e n o f i b r a t e p r o b u c o l a p r e p i t a n t t a d a l a f i l f e l o d i p i n e c a r v e d i l o l k e t o c o n a z o l e i b u p r o f e n v a l s a r t af i r l u k a s t t a d a l a f i l b r o m o c r i p t i n e a p r e p i t a n t f e n o f i b r a t e c a r v e d i l o l f e l o d i p i n e k e t o c o n a z o l e i b u p r o f e n v a l s a r t a n 1 Fig. 5 .
Fig. 5.The mean solubility values in the fasted (A) and fed (B) state in human intestinal fluids (HIF) from this study (Table4), as well as reported values in fasted (FaSSIF) and fed (FeSSIF) states of simulated intestinal fluids (Table5).A (-) indicates a higher (3-fold) solubility in simulated fluids than in HIF, and (+) indicates a higher (3-fold) solubility in HIF than in simulated fluids.

Fig. 6 .
Fig.6.Comparison of the mean drug solubility ratio in the fasted vs. fed (fed/ fasted) state human intestinal fluids (HIF) from this study (Table4) with reported ratios in fasted (FaSSIF) vs. fed (FeSSIF) state simulated intestinal fluids (Table5).The solid line shows the ideal agreement.The upper and lower dashed lines show where the simulated intestinal fluids overpredics (>3 fold) and underpredicts (<0.33 fold) the increase in fed/fasted HIF solubility, respectively.

Table 1
Some physicochemical properties and BCS class of the 17 drug compounds in this study.
5-n-octanol − water partition coefficient at pH 7.4/6.5,MMmolar mass, pKadissociation constant, Tmmelting point values.The properties of the confidential drugs are from the literature

Table 5
Reported mean (±SD) equilibrium solubility of the known model drugs in fasted and fed state simulated intestinal fluids (FaSSIF/FeSSIF).