Roller Compaction Scale-Up
The use of hydrophilic cellulose ether polymers such as hydroxypropyl methylcellulose (HPMC) in the controlled release (CR) of drugs has been well documented. Most investigations involved the use of cellulose ethers with direct-compression and wet-granulation technologies (1-8). Few studies of cellulose ether polymers were based on roll compaction techniques (9-14).
Roll compaction is a dry compaction-granulation process in which uniformly mixed powders are compressed between two counterrotating roll pairs to form a compressed sheet or ribbon that is then milled (granulated). The advantages of roll compaction technology in the pharmaceutical industry include a dry granulation system, high-volume production of granules, and good control of final particle bulk density and flow properties (15). A granulation step is sometimes needed to obtain adequate flow properties in CR formulations in which the HPMC polymer level is >20% of the final tablet weight. The formulation must flow evenly on high-speed tablet equipment to maintain uniform weight and drug content in each tablet. Because cellulose ether polymers are hydrophilic, water addition can make wet granulation challenging. Thus, a dry process that creates uniform powder flow and does not interfere with the final physical characteristics and drug release of the tablet could be useful.
Biopharmaceutic drug characterization (e.g., solubility and permeability) can facilitate CR efforts, because drug characterization allows for the consideration of drug substance influence on critical formulation and process variables. For example, poor drug solubility can reduce CR product release beyond anticipated CR formulation effects. A decrease in drug intestinal permeability also can narrow the acceptable range of critical formulation and process variables. The proposed Biopharmaceutics Classification System (BCS) considers drug solubility and permeability in gauging the propensity of a formulation to provide the same pharmacokinetic (and hence pharmacologic) profile that previous formulations have demonstrated (16). Although the BCS covers immediate-release (IR) products, this article extends BCS application to HPMC-based CR matrix theophylline tablets that have been manufactured by roll compaction.
Most previous studies of roll compaction and cellulose ethers have included laboratory experiments and in vitro performance measurements (9-14). Scaling up a pharmaceutical rollcompaction process can involve several issues and technologies (17). In any case, continued formulation development and scale-up to larger equipment typically would be performed independent of in vivo determinations. Integration of drug release and drug intestinal permeation data into early stages of CR product formulation development is not common. However, development time and cost can be reduced by combining in vitro drug-release data with information about the drug permeability across Caco-2 monolayers to predict human absorption kinetics at an early stage of product development.
This study involves the scale-up of a formulation containing HPMC and the model drug theophylline from laboratory to pilot plant using roll compaction. Research also was conducted to classify theophylline in the proposed BCS and to integrate these biopharmaceutic properties with HPMC formulation variables into a model to investigate in vivo performance of theophylline CR formulations containing HPMC. In addition, this work implements in vitro-in vivo correlation (IVIVC) from the laboratory to the pilot plant to predict in vivo bioequivalence among formulations developed under various process scales and speeds.
MATERIALS
Tablet preparation. The following materials were used as received: theophylline, USP, anhydrous powder >99% pure (BASF Corporation USA, Mount Olive, NJ); Methocel K4M Premium CR grade HPMC 2208, USP, 4000 mPa s (methoxyl = 19-24% [actual = 22.5%], hydroxypropoxyl = 7-12% [actual = 8.8%], viscosity = 3000-5600 cP [actual = 4447 cP]) (The Dow Chemical Company, Midland, MI); Fast Flo 316 lactose monohydrate, NF (Foremost Farms USA, Bamboo, WI); and magnesium stearate, impalpable powder, NF (Mallinckrodt Baker, Inc., Paris, KY).
Drug solubility, permeability, and predicted in vivo performance.
Caco-2 cells were obtained from ATCC (Rockville, MD). Radiolabeled 14C-mannitol was obtained from DuPont NEN (Boston, MA). Cell culture media components were purchased from Sigma Chemical Company (St. Louis, MO). Tissue-culturetreated, 10-cm plates and six-well cell-cluster dishes with polycarbonate Transwell filters (3-p,m pore size) were purchased from Corning-Costar (Cambridge, MA).
METHODS
Tablet preparation. The following materials were used as received: theophylline, USP, anhydrous powder >99% pure (BASF Corporation USA, Mount Olive, NJ); Methocel K4M Premium CR grade HPMC 2208, USP, 4000 mPa s (methoxyl = 19-24% [actual = 22.5%], hydroxypropoxyl = 7-12% [actual = 8.8%], viscosity = 3000-5600 cP [actual = 4447 cP]) (The Dow Chemical Company, Midland, MI); Fast Flo 316 lactose monohydrate, NF (Foremost Farms USA, Bamboo, WI); and magnesium stearate, impalpable powder, NF (Mallinckrodt Baker, Inc., Paris, KY).
Drug solubility, permeability, and predicted in vivo performance.
Caco-2 cells were obtained from ATCC (Rockville, MD). Radiolabeled 14C-mannitol was obtained from DuPont NEN (Boston, MA). Cell culture media components were purchased from Sigma Chemical Company (St. Louis, MO). Tissue-culturetreated, 10-cm plates and six-well cell-cluster dishes with polycarbonate Transwell filters (3-p,m pore size) were purchased from Corning-Costar (Cambridge, MA).
METHODS
Tablet preparation. Formulation. The tablet formulation included
• CR polymer: Methocel K4M Premium, USP (wt% = 30)
• filler excipient: Fast Flo 316 lactose (wt% = 19.75)
• model drug compound: theophylline (wt% = 50)
• lubricant: magnesium stearate (wt% = 0.25).
Mixing. For the laboratory experiments, powder mixes of 2 kg each were prepared in an 8-qt, twin-shell blender (PattersonKelley Company, East Stroudsburg, PA). For the pilot-plant experiments, mixes of 25 kg each were prepared in a 5-ft3, twinshell blender (Patterson-Kelley Company). Both blenders were equipped with an intensifier bar.
Polymer was the first ingredient added into the mixer, followed by drug, then lactose. The ingredients were mixed for 3.5 min with the intensifier bar turned off, then for 3 min with the bar turned on, and finally for 3.5 min with the intensifier bar turned off again, for a total mixing time of 10 min. A 10-min nitrogen purge was performed inside the blender when the powder ingredients were added and before each mix.
The lubricant was combined with a portion of the previously prepared mix and passed by hand through an 18-mesh wire screen. This mix was then added to the same twin-shell blender containing the remaining powder blend and mixed for an additional 2 min.
Roll-compaction sequence. Table I lists the processing parameters used in the various trials. Initial testing was conducted in a roller compactor (model TF-Mini, Vector Corporation, Marion, IA) to determine the optimal parameters for the compactiongranulation process. The compactor was equipped with concavoconvex compaction rolls (type DPS, diameter = 3.973 in. [100 mm], effective width = 0.984 in. [25 mm]) (see Figure 1) and a single-flight feed screw (type B) (18-20). Data were collected using LabVIEW software (National Instruments Corporation, Austin, TX).
The compacted ribbons were milled with a Comil grinder (model 197S, Quadro Engineering, Waterloo, ON, Canada) equipped with a round-hole, grater-type screen (2A-D62G03123139) and a rotating impeller (2A-1601-173). The objective of the milling process was to produce a free-flowing granulation that contained -25% or less fines. Fines were defined as particles sized -- 140 p,m (100-mesh US Standard). Optimal results were obtained using a roll speed of 6 rpm, a screw feeder speed of 17.9 rpm, and a roll force of 3 tons (1250 psi). These settings resulted in a linear roll velocity of 74.2 in./min, a force-per-linear-inch of 3.1 tons per inch of roll width (i.e., 3 tons per 0.984 in. [25 mm] of roll width), and a ratio of screw speed to roll speed of 3:1 (i.e., 17.9 rpm screw speed to 6 rpm roll speed).
These parameters were then scaled directly to another roller compactor (model TF-156, Vector Corporation). The scale-up experimental protocol in this study was established using FDA's Guidance for Industry SUPAC-MR: Modified-Release Solid Oral Dosage Forms, Scale-Up and Postapproval Changes: Chemistry, Manufacturing, and Controls; In Trtro Dissolution Testing and In Vivo Bioequivalence Documentation (September 1997). To determine that the particle-size reduction equipment used in this study was of the same class and subclass, Vector Corporation referred to FDA's Guidance for Industry SUPACIRlMR: Immediate-Release and Modified-Release Solid Oral Dosage Forms, Manufacturing Equipment Addendum (draft guidance, April 1998).
Figure 2 shows the TF-Mini and the TF-156 roll compactors used in this study. Like the TF-Mini model, the TF-156 unit was equipped with concavo-convex compaction rolls and a singleflight feed screw. The rolls were 150 mm in diameter with an effective width of 44 mm (see Figure 1). An integral part of this unit is a rotating-bar granulator located directly beneath the compaction rolls. The roll speed of the TF-56 model was scaled to achieve the same linear velocity (74.2 in./min) provided by the TF-Mini model. Using this setting maintained a comparable dwell time for material in the compaction zone. Feed-screw speed and roll speed have been shown to have the most significant effect on granulation characteristics (21). For the initial TF156 trial (Trial 2), a roll speed of 4 rpm was used, which provided a 74.2 in./min linear roll velocity. The roll force was scaled to 5.6 tons, which provided a force-per-linear-inch approximately equal to that provided by the TF-Mini model (i.e., 3.1 ton/in. roll width). The screw speed was initially set at 12 rpm, which maintained the same ratio of screw speed to roll speed that was used for the smaller unit. However, this appeared to overfeed the rolls. The screw feeder speed was then adjusted to achieve an adequate delivery of powder to the compaction zone.
The feed of the theophylline blend was optimized at a screw speed of 5.2 rpm. This setting established a baseline ratio of screw speed to roll speed of 1.3:1 (5.2 to 4.0 rpm), which was used for the remaining trials. For the initial scale-up (Trial 2), the milling method (rotating impeller) was the same as that used in the TF-Mini laboratory trial (Trial 1). In all subsequent trials, the rotating bar granulator supplied with the unit was used as the method of milling the compacted ribbon. The gravity feed of the compacted ribbon to the granulator provided hands-off processing.
To determine the effect of compaction dwell time on the quality of the compacted ribbon, the roll speed was first increased by a factor of 2 (8 rpm) in Trials 7 and 8 and then by a factor of 4 (16 rpm) in Trials 9 and 10. These settings resulted in linear roll velocities of 148.4 and 296.8 in./min, respectively. A ratio of screw speed to roll speed of 1.3:1 was maintained for Trials 7-10. To determine the robustness of the process, high and low roll forces were used. For Trials 5, 8, and 10, the force was increased to 6.6 tons (3.8 ton/in.). For Trial 6, a reduced force of 4.6 tons (2.7 ton/in.) was used.
Tablet pressing. Tablets were prepared from the milled granulations on an instrumented 16-station Manesty Betapress (Thomas Engineering, Inc., Hoffman Estates, IL) modified by D&F Farmaceutica Service (Martin, MI) and equipped with 0.27 X 0.49 in. oval tooling. Tablet tooling similar to that of an abbreviated new drug application (ANDA) 200-mg theophylline product was used to minimize surface area effects during drug-release testing. The tablet press data acquisition and analysis system was provided by Dateppli, Inc. (Kalamazoo, MI). A final compression force of 17.8 kN was applied during the preparation of the tablets from each of the original runs (direct-compression, laboratory, and pilot-plant roller compaction runs). A target tablet weight of 400 mg ± 3% was used.
Roll-compacted ribbon/granule testing. Samples of the surface and fractured cross sections of the roll-compacted ribbons were mounted on a scanning electron microscope (SEM) sample stub with conductive silver paint. Each sample was coated with a conductive and contrast-enhancing layer of gold-palladium. The samples were examined with a field emission gun SEM (model DS130F, Topcon Instruments, Paramus, NJ) using secondary imaging at 5 keV and a working distance of 15 mm. Observations were made and digital images were recorded at a 500 X magnification.
The milled granules were analyzed for density, percent compressibility (Carr's index, CI), flow, and particle-size distribution (22). The density of a 40-g sample of each granulation was tested using a tap density tester (Vanderkamp model 10700,VanKel Industries, Edison, NJ). The volume was measured before tapping (apparent density) and after 500 taps (tap density). Percent compressibility was calculated using the following equation.
Where T is the tapped (packed) density and B is the bulk (apparent) density (both in g/cm3).
A 40-g sample of each granulation was shaken for 5 min on a sieve shaker (RoTap model B, W.S. Tyler, Gastonia, NC) equipped with a series of five screens and a pan. The amount of material retained on each screen size was measured, and a particle-size distribution was calculated.
Tablet property testing. Tablets were tested for friability, thickness, hardness, and drug release. The friability of 20 randomly chosen tablets from each granulation trial was determined by tumbling the tablets for 6 min in a friabilator (Vanderkamp model 10801, VanKel Industries) and then measuring the percent weight loss. The hardness or crushing strength of 20 randomly chosen tablets from each granulation trial and 5 from the ANDA 200-mg commercial theophylline product were tested using a hardness tester (model HT500, Key International, Englishtown, NJ). Tablets were placed on their sides and crushed end-to-end. Tablet thickness was measured with a caliper (Absolute Digimatic series 500, Mitutoyo Corp., Japan) for 10 of the 20 randomly chosen tablets. The weight of each of the 20 tablets was measured to ensure weight variation was within acceptable USP limits (23).
Drug-release testing of six tablet samples from each variable run was performed using a dissolution system (model 2100, Distek, Monmouth Junction, NJ). Testing was performed in accordance with USP 23/NF 18, Official Monographs, Theophylline Extended-Release Capsules, Test 1 (for products labeled for dosing every 12 h) (23). The USP Apparatus 2 (paddles) method was used at an agitation speed of 50 rpm. Three-prong clips were used to prevent tablets from floating on the surface of the dissolution media during testing. The dissolution media consisted of 900 ml, of pH 1.2 simulated gastric fluid (without pepsin) for the first hour and pH 6.0 phosphate buffer thereafter. Data were acquired via a diode-array spectrophotometer (model 8452A, Hewlett-Packard Company, Valley Forge, PA). Drug-release profiles were generated at 37.5 °C, with detection at 268 nm for theophylline.
Drug-release profile testing. Drug-release profiles were compared using equations for the fl and f2 metrics (24,25). The value of the ft metric represents the approximate percent error between two curves such as those of a test formulation and a reference formulation. The value off, increases in proportion to the dissimilarity between two profiles. Thus one formulation is chosen as the reference formulation, and the remaining formulations are designated as test formulations. The f2 metric is the similarity factor, and values of f2 between 50 and 100 indicate profile similarity. An f2 value <50 suggests that the two drug-release curves differ by at least 10%. In Equations 2a and 2b, Rt is the reference profile, Tl is the test profile, n is the number of data points collected, and wt is an optional weight factor. Only f2 testing was used in this study.
Drug solubility (cs) of anhydrous theophylline was determined at 37 °C between pH 1.2 and 8.0 using the conventional direct approach (i.e., phase-solubility method) (26). The following media were used: simulated gastric fluid, USP, without pepsin at pH 1.2; acid phthalate buffer, USP, at pH 4.0; and simulated intestinal fluid, USP, without pancreatin at pH 6.0, 6.8, and 8.0. In the phase-solubility approach, a sample will yield either a solution when low amounts of drug are added or a suspension when excess drug is added. Undissolved theophylline was removed by filtration through a 0.22-p,m filter. Dissolved drug rug concentration, and data were analyzed via linear regression to estimate cs. Freshly prepared control samples with known drug solution concentration were analyzed for chemical instability assessment. Theophylline was stable (<5% loss) at all pH values during the entire study.
Transition to monohydrate. The transition of anhydrous theophylline to the monohydrate needle habit was observed through a polarized-light microscope (Nikon Eclipse E600 POL equipped with a Microflex II-3 automatic camera system, Nikon Inc., Melville, NJ). Anhydrous theophylline was placed in water on a microscope slide at room temperature. Photomicrographs were taken at 0, 3, and 10 min after water contact.
Permeability. Caco-2 cells were cultured using a previously developed 4-day growth protocol (27). The protocol involved seeding the Caco-2 cells at a seeding density of 1.00 X 106 cells per 4.71 cmz. Cell monolayers were grown in a 1:1 mixture of DMEM/F-12 with 2% iron-supplemented calf serum and additional cell-culture constituents. The medium was changed after 48 h, then daily for the remainder of the culture period. These cells were cultured for 4 days at 37 °C, 90% relative humidity, and 5% CO2, then used for permeability studies.
Theophylline permeability studies were conducted in Hank's balanced salt solution (HBSS) containing 10 mM HEPES buffer at 37 °C and 50 oscillations per min. Monolayer integrity was monitored using transepithelial electrical resistance in HBSS (resistance >500 fl cm2 at ambient room temperature after subtracting a filter resistance of 610 fl cm2) and 14C-mannitol permeability. Metoprolol tartrate was used as the high-permeability reference compound. (Drug permeability greater than that of metoprolol is classified as high permeability.) Transport studies were conducted in both the apical-to-basolateral (AP-BL) and the basolateral-to-apical (BL-AP) directions using 1.0 and 0.10 mM concentrations of theophylline. Drug was quantified using HPLC analysis. A scintillation counter was used to quantify the amount of 14C-mannitol. In each study, mass balance was >95%.
IVIVC predicted in vivo tablet performance. A convolutionc approach using WinNonlin to IVIVC was used. Figure 3 illustrates the approach to integrate drug dissolution, theophylline intestinal permeability, and theophylline pharmacokinetics into an IVIVC model for theophylline HPMC CR matrix tablets. First, a model representation of each formulation's observed drug-release profile was sought. Model selection can influence IVIVC analysis acceptability (28). Several models were considered, including the modified Higuchi equation:
Model fits to each formulation were obtained using WinNonlin. The Akaike Information Criterion (AIC) was used to select the best-fitting model from several competing models for all formulations (29). The best-fit model will have the lowest AIC value. The best-fit model was then used in subsequent determinations of in vivo tablet performance. Second, a model representation of theophylline intestinal permeation was devised using the Caco-2 permeability data for theophylline where kp is the first-order permeation coefficient, A is the absorptive surface area of the gastrointestinal tract, V is the volume of the gastrointestinal fluid, and P is the Caco-2 permeability. Equation 4 appears to be a reasonable basis on which to scale permeability data to human drug absorption kinetics (30).
Finally, theophylline pharmacokinetic parameters were taken from Pollack et al. (31). Parameter values were as follows: volume of distribution = 29.8 L, elimination rate constant = 0.102/h, and time lag = 0.508 h.
Predicted in vivo tablet performance was conducted through convolution of the drug-release profiles. Predicted relative differences in Cm. andAUCi„f were examined in accordance with FDA's IVIVC guidance (32). Bioequivalence was anticipated if the average absolute percent prediction difference (% PD) was 15% for Cmax and AUCinf>.
RESULTS AND DISCUSSION
Roll-compacted ribbon/granule testing. Figure 4 shows SEM photographs (magnification = 500X) of the external surfaces of rollcompacted ribbons manufactured under laboratory and pilotplant conditions. The sample from Trial 1 (laboratory) exhibited the smoothest surface of all samples tested, having only a few small cracks and voids and a relatively clean fractured surface. The sample from Trial 4 (pilot-plant scale, 4-rpm roll speed) exhibited a slightly rougher surface than the sample from Trial 1 with more small cracks and voids present on the surface. The sample from Trial 8 (pilot-plant scale, 8-rpm roll speed) exhibited a considerably rougher surface than the latter two samples, with an increased number of sizable cracks and voids. Numerous voids contained what appeared to be fibers and fiber debris. Some loose material also was observed on the surface of the sample's cross section. The sample from Trial 10 (pilot-plant scale, 16-rpm roll speed) exhibited large, clearly visible cracks and voids, which appeared to contain loose material.
Table II lists the properties of the original formulation (not roll compacted) and the granules from the milled, roll-compacted formulation. In general the combination of roll compaction and milling increased both the bulk density and the tap density of the formulation. Small differences existed between the values of the bulk and tap densities of Trial 1 (laboratory) and those of its direct pilot-plant scale-up, Trial 2. Therefore, there does not appear to be a correlation between pilot-plant equipment conditions and density.
The CI value is based on the values for the bulk and tap densities and indirectly represents the flowability of a powder mass. A low CI value indicates that the material flows easily. All of the trials - laboratory and pilot-plant scale - exhibited improved flowability in comparison with the non-roll-compacted original formulation.
Figure 5 shows the particle-size distribution for Trial 1 (laboratory) and Trials 2, 4, and 10 (pilot-plant). Trials 1 and 2 (both of which involved a rotating impeller) had a greater number of large particles than did Trials 4 and 10 (both of which involved rotating bar granulator). In general formulations that contained a greater number of large particles also had higher densities.
Tablet property testing. Table III lists the physical properties of the commercial ANDA tablets, the original formulation (directcompression sample), and the samples that had been milled and roll compacted (Trials 1-10). Usually tablet friability relates inversely to tablet hardness (crushing strength) - the lower the value for tablet hardness, the higher the value for tablet friability. In this study, the friability of all samples was < 1 % (data not shown). This suggests that all tablets exhibited acceptable tablet physical characteristics. All tablets exhibited strong physical properties with very little chipping, no breaking, and no physical defects such as capping or lamination.
All of the samples exhibited weight variation that was -_ 1.2% of the 400-mg ideal tablet weight, well within USP guidelines. All tablets produced by roll compaction exhibited higher values in tablet hardness than the those of the ANDA product and lower values in hardness than tablets produced by direct compression. The lowest values in tablet hardness were observed in Trial 1 (laboratory scale). However, tablets from Trial 2 (direct scale-up) exhibited hardness values within 4% of those from Trial 1. These results agree with the results of previous studies (33,34). In addition there did not appear to be any correlation between tablet thickness and hardness.
Drug-release testing. Figure 6 shows the drug-release profiles of the ANDA 200-mg theophylline product, the tablets produced by direct compression, and the tablets from the laboratory and pilot-plant trials. (Trial 3 was not run because of its similarity to Trial 4.) Table IV compares the f2 metric values using the ANDA product as a reference and Trial 1 (laboratory) as a reference. All f2 values were between 50 and 100, indicating that tablets from Trials 2 and 4-10 provided a drug release similar to those of the ANDA product and tablets from Trial 1. Despite differences in tablet hardness and slight differences in other tablet properties, there were no significant differences in the rate of drug release among the ANDA product, the tablets produced by direct compression, and the tablets produced using laboratory or pilot-plant roll compaction of the original formulation. Drug solubility, permeability, and predicted in vivo performance. Solubility. Table V lists the apparent solubilities and the dose numbers (35) for anhydrous theophylline at pH 1.2, 4.0, 6.0, 6.8, and 8.0. Anhydrous theophylline solubility was classified as high because solubility at each pH exceeded 1.80 mg/mL. This threshold value was obtained by dividing the maximum available dose (450 mg) by 250 mL, in accordance with the BCS (16). The imidazole group of theophylline possesses both proton-accepting and proton-donating capabilities, with pKa values of 8.77 and 2.5, respectively (36). Hence theophylline is essentially nonionized between pH 4 and 6.8. This ionization profile agrees with the pH-independent solubility profile in Table V. The modest increase in solubility from pH 6.8 to 8.0 can be attributed to ionization of the acidic imidazole. This solubility increase at pH 8 agrees with observations from Ledwidge and Corrigan (37).
The dose number at each pH is 0.05, much less than 1, indicating that dissolution alone will not be a major kinetic barrier to drug release. The dose number takes into account the mass of the drug for complete dissolution, the effective volume into which the drug will dissolve, and the drug's solubility (35). These data indicate theophylline release will be determined by the HPMC matrix, with no resistance caused by theophylline drug substance solubility alone.
Transition to monohydrate. Figure 7 shows the rapid conversion from anhydrous theophylline particles to theophylline monohydrate needles. At 0 min, under polarized light, anhydrous theophylline particles appear as small granules against a black background. Anisotropic substances interact with polarized light and will appear colored between crossed polarizers. At 3 min after water exposure, a substantial number of theophylline monohydrate needles exist. At 10 min after water exposure, nearly all of the particles are theophylline monohydrate needles. Color variations result from needle thickness variations.
Theophylline monohydrate develops from solutions of anhydrous theophylline (3611). This conversion occurs soon after contact with water (37) and was observed in this study using polarized-light microscopy. De Smidt et al. observed conversion of anhydrous theophylline to theophylline monohydrate -v 100 s after the tablets were exposed to buffer solution (38). Needles of theophylline monohydrate formed within 3 min of water contact. Hence, the measured solubility was strictly that of theophylline monohydrate. The theoretical value of the solubility of the anhydrate is higher than that of the monohydrate.
Subjecting anhydrous theophylline to wet-granulation methods can result in theophylline monohydrate (42). In the presence of microcrystalline cellulose, nearly all anhydrous theophylline was converted to theophylline monohydrate after wet granulation. At higher levels of granulating fluid theophylline release from 60% drug-loaded pellets occurred at a significantly slower rate. In another study anhydrous theophylline was converted to its monohydrate by exposure to 98% relative humidity (43). Theophylline monohydrate was then dried at room temperature under reduced pressure to form a metastable anhydrous polymorph. In tablets the metastable anhydrous polymorph yielded the stable anhydrous theophylline within 10 days of storage. This recrystallization was accompanied by a pronounced decrease in drug-release rate. Permeability. Table VI lists AP-BL and BL-AP permeability values for theophylline at 1.0 and 0.10 mM. Theophylline permeability is -20% greater than metoprolol permeability in the AP-BL direction and 50% greater than the metoprolol permeability in the BL-AP direction. Similar permeabilities in both directions for 1.0 and 0.10 mM indicate theophylline transport occurred by passive diffusion.
Solubility and permeability results suggest that theophylline is a Class I drug (high solubility, high permeability) in the BCS. For theophylline formulated into HPMC CR matrices, drug solubility is not expected to be a major barrier to drug release. Rather, drug diffusion and erosion will be the dominant releaserate-limiting step. High drug permeability, such as that predicted for theophylline, is a necessary attribute for CR delivery. IVIVC predictions of in vivo tablet performance. A modified Higuchi equation served as the best-fit dissolution model. For clarity, fits were not shown in Figure 8. However in all cases, the model provided very close agreement between fitted and observed data (r2 >0.999, deviations -_ 1.36% dissolved).
Figure 9 shows the predicted plasma concentration profiles for the formulations in Trials 1, 4, 8, and 10. Table VII lists the Cmax AUCinf and % PD values of each formulation in comparison with the values for Trial 4. In all cases the absolute % PD was < 15%. Hence as shown in Figure 3, drug release, theophylline permeability through Caco-2 monolayers, and a onecompartment open model were applied to predict plasma profiles from dissolution data. This approach, which can be performed at an early stage in drug development, indicates that roll-compacted HPMC CR matrices will provide similar product performance for theophylline, a Class I drug, even for the relatively wide formulation and processing ranges used in this study.
CONCLUSION> This study incorporated drug release, theophylline permeability, and theophylline pharmacokinetics data into early stages of formulation and process development. This approach may reduce development time and costs and may provide the basis for a more rigorous, validated IVIVC for drug-release specification justification at later stages of development (32). Results showed that although roll-compaction scale and equipment conditions caused slight variations in the tablets' physical properties, these differences did not affect in vitro drug release. The HPMC CR formulation exhibited robustness both during the roll compaction process and tablet testing.
Anhydrous theophylline was determined to be a Class I drug in the BCS (i.e., high solubility, high permeability). Drugrelease profiles of all formulations best fit a modified Higuchi equation. An IVIVC model, which used drug-release and BCS data, predicted bioequivalence among laboratory, pilot-plant, and two-variant pilot-plant formulations, based on predicted values of Cmax and AUC. These results suggest that HPMCbased matrix tablets, manufactured by roll compaction under various conditions, will provide similar product performance for Class I drugs, even for relatively wide formulation and processing ranges.
ACKNOWLEDGMENTS
Vector Corporation would like to acknowledge the following individuals from The Dow Chemical Company: Kerry Pacholke, senior R&D technologist, and Christy Fischer, student technologist, for assistance with the laboratory testing; David Williams for assistance in generating the SEM images used in this study; and Jeremy Brand for assistance in digital print preparation. Vector Corporation also would like to thank Leo Lucisano of Glaxo Wellcome, Inc., for his SUPAC guidance expertise and Ann Birch of ediTech for her assistance with the manuscript.
<%=company%>