Dimensional orthogonal imaging of laminar fluid flow across API surface : Insight into dosage concentration inside GI Lumen and permeability

Historically, dissolution devices have been performing measurements away from the solid-liquid interface and then drawing conclusions. Simply relocating measurements to this interface using a 2 dimensional imager, recording absorbance intensity over time of laminar flow across the sample surface, opens significant insight about the concentration gradient at the surface, in-situ surface pH, and unique insight of bioavailability. Having this information early in development provides formulators with better insight into optimum form.


Introduction
Dissolution of API material from a surface is an important characteristic measurement used in conjunction with solubility to determine an appropriate dosage. Intimate understanding of the dissolution process at the solid liquid interface is not possible with traditional equipment which leaves many problems with product development, particularly poorly soluble compounds. This development cycle quite often suffers from too little information too late. Surface imaging of a laminar flow-through cell offers a tremendous amount of information for development teams from synthesis through first in trials. First introduced in 2008 1 , UV imaging coupled with laminar flow is able to generate 4dimensional data from absorbance intensity over a 2-dimensional area over time. When this technique is applied to dissolution 11 it offers incredible information about the solid-liquid interface such as static diffusion rate, boundary layer thickness, boundary layer slope, dynamic solubility, surface solubility, in-situ surface pH, and kinematic viscosity. The minute sample and buffer amounts of 3 -10 mg and 5 -10 ml, respectively, analyzed over 10-15 minutes make this incredibly attractive to early stage development 3,4,5 . Likewise, pharmacokinetics will appreciate a specially designed micro flow cell (1/10 the diameter of the small intestine) that maintains laminar conditions over a volumetric and linear velocity range equivalent to published Human SI rates 4,6,22 . With the Reynolds number and flow rate similarities, it is not much of a stretch to consider the small sample surface area as 1mm particle API situated on the villi surface of the Lumen. Such rich information from small amounts of sample will be appreciated by all researchers.

Results and Discussion
Theoretical understanding of fluid dynamics of laminar flow through devices 9,10,16,17,21 has been available for quite some time, and flow through devices are readily used today, but not extensively applied 2 . Laminar flowing devices specialized for small, compressed API powders were investigated thoroughly as early as 1975 for use as an alternative to measure the dissolution process 9,10 . Since then, very little progress has been made until recently 1,17 . A laminar flow through device has many benefits that seem to be under appreciated, such as continuous flowing regimes similar to physiological conditions with low Reynolds number, steady state conditions, and continuous supply of buffer.
Convective diffusion models and evidence have been thoroughly presented, but overall the technique has failed to successfully gain support as a viable research tool. One possible explanation, which has a similarity to traditional apparatuses, would be disconnected measurements, or analytical measurements at a large distance away from the original reaction leaving a hole for explanations of the detailed process. The obvious cure for this is to move the analytical measurement to the actual solidliquid interface, which has recently been successfully demonstrated 1,11 .
Convective diffusion of a laminar flow through device was thoroughly characterized by Nelson and Shah. The predicted diffusion rate could be derived based on parameters of the compound, fluid flow, and device dimensions. This is seen in Eq. 1 where DR is diffusion rate, D the diffusion coefficient, S the maximum solubility, α represents the shear rate, and w, L the cell width and length respectively. Whereas in a 2-dimensional imager a mass flux is measured to provide the intrinsic dissolution rates (IDR) based on measurements in a downstream detection zone using Eq. 2. Briefly, a measurement zone defined as a fixed width vertical column containing hundreds of rows of pixels, records material transported downstream by each row of pixels. The mass in each pixel is obtained from the molecular weight and extinction coefficient as applied through Beer's law. The mass flux is corrected by pixel row away from the surface for the velocity in that plane using Eq. 3 to accommodate faster velocities in the central flow stream and near zero flow at the surface. Nominally, Q is flow rate, h the cell height, w the cell width and Z the pixel row height from surface.
Fluid flow into the flow cell, Fig 2b, is controlled through a high accuracy, variable rate dual position syringe pump connected to a heating element to raise the temperature of the buffer prior to entering the flow cell. The elliptical shape of the flow cell, seen in Fig 2 (c) has been carefully designed to provide a laminar flow throughout the entire cell. Internal volume of the flow cell is approximately 0.5 ml with internal dimensions at the centre of 4.0 mm wide x 3.5 mm high. This cross sectional area was chosen based on dimensions of the camera chip, but also offers a 1/10 model of the small intestine. Fluid pumping can also be managed by a standard HPLC pump where it can be realized that a continuous flow pH gradient can be put through the cell using the same sample to collect dissolution rate data at several pH in one experiment, which has the benefit of easily mimicking the pH change of the GI tract. Further to this, because the cell volume is small, rapid flow rate changes can be easily managed with the ability to go from high velocity (2.0 ml/min) down to no velocity (0.0 ml/min) instantaneously without any ill effects.
Many additional benefits are available from a small flow through system. For a total experiment of 12.5 hrs a mere 150 ml of buffer would be used, which is very beneficial when using biorelevant media. From Table 1 a flow rate of 0.2 ml/min corresponds to nominal published physiological linear velocities, 1.3 -1.5 cm/min 22 (or 0.25 mm/sec) which is approximately 18 sec to move across the face of the detector. Further, effluent collected for 5 minutes, at this flow rate, would offer 1 ml of solution, which is a manageable volume for direct use with permeability systems such as PAMPA or Caco-2. The effluent concentration can be estimated by dividing the measured IDR by the flow rate to arrive at a concentration (mg/ml). This value must be corrected for the ratio of sample area to cell cross sectional area as in Eq. 4. This concentration is the solubility in a dynamic system, or the dynamic solubility (dS) and its relationship to IDR and flow rate are seen in example of Fig. 1.

4-Dimensional data from 2-Dimensional UV detector
A parallel pixel array, which has many similar properties to a digital camera, offers an impressive amount of data when trained on a reaction zone such as a solid-liquid interface. Resolution and orientation of image relative to the reaction zone are explained in Fig 2. With each pixel acting as an individual UV detector, an unparalleled level of temporal and high resolution spatial data directly of the solid-liquid interface is now available. Relocating measurement of the dissolution process offers insight into previously hypothesized surface events such as surface pH, boundary layer thickness, and concentration gradient. However, further data can be collected on static diffusion rate, boundary layer slope, boundary layer diffusion rate, surface concentration, dynamic solubility, and surface changes from swelling or gelling.
The 4 mm optical path length offers a good range of sensitivity, especially for very poorly soluble compounds. Because the highest concentration achievable from a poorly soluble compound is directly at the surface (before dissolved material has been allowed to dilute), detection limits should be the highest here. Indeed, experience with a variety of very poorly soluble API compounds shows that those unable to be detected have a physical change at the surface. By selecting the wavelength corresponding to the chromophore of interest, concentration can be tracked in the immediate vicinity of the sample. This advantage is used for measurement of surface pH described below.
Ketoprofen, a poorly soluble compound investigated often 3,5,7,8 , as displayed in Fig 3 shows the sensitivity and positional accuracy available. Using preferred flow rates from above, the 0.1 and 0.3 ml/min profile lines in Fig. 9 show the distance to which this compound will be carried into the flow stream and the concentration at the surface. These profile plots are created from any of three vertical measurement wands that can be placed along the X axis relative to the sample surface as desired. Thus, profiles can be found directly over the center or further downstream.  Actual false colored image shown next to viewing area. The Z axis is away from the surface, X left to right and Y axis as depth. Horizontal viewing area of the chip is fixed at 9 mm, but the viewing height can be adjusted to full cell or less depending on data collection rate and required resolution.

in-situ Surface pH measurement
It has been known that the pKa of compounds can influence the bulk pH 5,7,13,14 . Equally so, it has been known that the boundary layer is the driving force behind the dissolution rate. From a practical point of view, the more material transferred from the surface to the bulk solution the higher the absorbance. Thus determining either the best salt form to control the in-situ surface pH or excipients to increase the boundary layer will improve the dissolution rate of the API in question. Herein lays the origin of designing supersaturated solutions of poorly soluble compounds to influence the rate of material leaving the surface.
Measurement of the surface pH depends on the difference between the compound pKa and buffer pH and the matching of the pKa to the indicator dye. Less than 2 pH units are difficult to measure, and optimally the indicator dye range should span the pKa. In the case of Ketoprofen, Methyl Red (MR) diluted solution was prepared using the pH 6.8 buffer. All pH indicator dyes change color, and thus absorbance, upon pH change. Fig 4 shows a plot of absorbance vs. pH change for MR. The dilute dye solution is pumped through the cell and the baseline is established with the dye in place.
A calibration of absorbance vs. pH was made previously to coordinate pixel intensity with pH. Previous estimate of surface pH of Ketoprofen has been made 7 using an independent measurement of a saturated bulk solution. The results obtained using a pH indicating dye is in good agreement, Fig 5. A similar experiment was conducted on Atenolol, a rapidly soluble compound with a pKa of 9. It is known that compounds such as this can alter the bulk solution pH. In this instance, with relatively low volumetric flow (0.3 ml/min or 2 cm/min) the density of the high pH solution was too heavy to remove out of the flow cell and was pulled down with gravity, Fig 6. This raises questions about the concentration at the surface. Similar high pKa compounds such as Vancomycin exhibit similar properties and performance seen in

Boundary layer
Understanding the role of the boundary layer as it applies to both the API and Lumenal surface is vital to understanding oral absorption. To date this has been very difficult to measure and understand from both the disconnected measurement location and time delay from reaction to measurement. Currently poorly soluble compounds receive a great deal of interest in order to enhance or control the boundary layer 13,18,20 Insight into the values and conditions at the surface can aid greatly in determining not only which form of API, but also solving problems associated with current product and batches.
A single data image, such as Fig 8, contains a wealth of information that can be found in the concentration gradient above the surface boundary. Three distinct values can be made for any compound: Surface concentration, slope of concentration gradient, and rate at which material leaves the gradient.
The surface concentration is easily found from the average intensity within a defined region of interest (ROI). The volume is the area of the ROI times the path length, and once absorbance value is converted to mass an absolute concentration is known. The slope of the concentration gradient is indicative of ease to which the dissolved material moves into the bulk stream. A more horizontal slope is an easy transition whereas a more vertical slope defines a sharp contrast. Finally the rate of material loss can be found from the difference between two vertical measurement wands for a chosen concentration. Using the increase in vertical height of concentration Z and the distance between the two wands creates a triangle to which the hypotenuse is the rate loss per linear distance.
Viewing the gradient as a function of flow rate, as in Fig 9, will reveal the extent to which a certain concentration can move into the bulk flow stream. Further to this, the area under the curve (AUC) is an additional differentiation value between API forms. Work has been carried on stability of the laminar dissolution steady state by comparing differences in AUC over several consecutive time points 1 min apart, which revealed reliable reproducibility.  The static diffusion rate can be found from the spatial difference of profiles between two time points to offer distance/time (mm/sec).

GI model
The gap between knowledge about physiological conditions, as they relate to dissolution, and dissolution apparatus improvement continues to widen each year leaving researchers with many unanswered problems 15 . The hope has long been for more knowledge, particularly towards predicting in-vivo results earlier 14 . Simulating the GI is complex and not easy. However, a steady state and reproducible model is an excellent starting point.
Starting with the premise that the current flow cell is a 1/10 model of the small intestine and that identical linear velocities are easily obtained, it is a small stretch to imagine the flow cell as a section of the intestine with a camera positioned to look through the wall at the internal flow characteristics. This would resemble the view seen in Fig 9 and 10 where Fig. 9 depicts integration of the sample cup with the Lumenal wall. Using the advantage of 2-dimensional data from an area imaging device, it can be seen how convective diffusion carries a finite concentration downstream and from the available measurements described above, what the concentration and pH would be at the Lumen surface in this scenario. Of course peristaltic mixing inside the SI ensures equal mixing of contents to provide maximum Villi exposure, but the measurements can be applied in the opposite direction. If a particle was in the middle of the flow stream, static diffusion would take too long to travel towards the Villi surface (VS), therefore mixing would be required to bring the particle closer to the VS, which is similar to the original premise. Regardless of particle location, the in-situ surface pH and surface concentration values would remain the same.
In the case of the particle dissolved upstream, a concentration plug entering the observation zone would show diminishing concentration at the surface with higher flow rates as previously explained. 19 Data images at different flow rates of identical flow injections of Caffeine, Fig 11, reveal the thickness of the aqueous boundary layer 18 and concentration entering this zone. In this case, for a discrete finite volume of plug entering, the 2-dimensional data can show concentration at the surface over time, which when applied to the total internal surface area can provide a measure of available concentration rate for a given inlet concentration (dosage). This model system is novel and can offer significant advantage if correlation with in-vivo results can be realized as all the data is collected simultaneously with the normal IDR.   (19). Shown are the relative values for flow in, concentration in/out and boundary layer thickness, which is a function of flow rate. Higher rates will have a thicker layer than slow. Figure 11. (a) Injections of caffeine showing how radial distribution is affected at different flow rates. Worth noting is the varying thickness of the stagnant layer seen in all images, which corresponds to δ boundary seen above.