Nipte.org

Formulation of a gabapentin drug degradation model that combines manufacturing and storage stress variables T. Radaduen, S. Stamatis, H. Q. Nguyen, Z. Zong, L. E. Kirsch National Institute for Pharmaceutical Technology and Education and The University of Iowa The objective of these studies was to formulate a drug degradation model that incorporated environmental storage and manufacturing stress factors and to devise an approach to apply the model in manufacturing design space and stability risk assessment applications. The effect of manufacturing stress on drug degradation kinetics were determined by co-milling gabapentin with 6.5 %w/w hydroxypropyl cellulose (HPC) in a Fritsch planetary micro mill Pulverisette 7 for 15 and 30 minutes to induce crystal defects under mechanical stress. Then the gabapentin/HPC mixtures were stored at 40-60 ˚C and 5-30% relative humidity. The rate of gabapentin-lactam formation was measured by HPLC. An irreversible two step autocatalytic reaction scheme was fit using Bayesian and Markov Chain Monte Carlo methods. Marginal posterior distributions show relatively high amounts of parameter correlation, but the model adequately described the data. Linear correlation between the predicted initial rate (at 50 °C and 5 % RH) and estimated level of manufacturing-stress damaged intermediate concentration was used to provide a proportionality constant for incorporation into design space and risk assessment models. A simplified model for time-dependent formation of lactam under typical shelf-life storage conditions (Dry/RT, Dry/40⁰C, 30%RH/RT and 30%RH/40⁰C) was also developed to predict lactam formation during long-term stability. Assessment of manufacturing stress in long-term stability of gabapentin tablets Hong-Shian Huang, P. Ngeacharernkul, S. Stamatis, L. E. Kirsch National Institute for Pharmaceutical Technology and Education and The University of Iowa The object of these studies was to use long-term stability results to validate a gabapentin degradation model that incorporated both environment storage and manufacturing stress factors by comparing model predicted and observed time-dependent lactam levels. In addition these studies were used to illustrate a procedure for formulating and validating a risk assessment model that predicted the probability of stability failure based on manufacturing conditions. Long-term stability data were obtained as follows. Selected tablet samples from pilot batches (prepared by the NIPTE unit operations team) were stored for six months in 20-mL glass vials at four different controlled conditions protected from light: Dry/RT, Dry/40⁰C, 30%RH/RT and 30%RH/40⁰C. Samples of one tablet or 40-60 mg of dried granules or final blends were removed from storage at different time points for extraction by water, followed by HPLC testing of gabapentin and its lactam content. The HPLC analysis used Waters µBondapak CN-RP, 3.9x300 mm column with flow rate at 1.0 mL/min and detection at 210 nm. The % lactam was calculated on molar basis. Degradation and risk assessment model predictions were generated using initial lactam measurement, initial rate of lactam formation (at 50 °C and 5 % RH) or manufacturing control parameters based on previously described models. Statistically-significant correlations demonstrated the applicability of this approach for incorporating stability risk assessment into drug manufacturing control. Role of Excipients in the Solid State Lactamization of Gabapentin Radaduen Tinmanee, Zhixin Zong, Lee E. Kirsch National Institute for Pharmaceutical Technology and Education and The University of Iowa Solid state degradation can be affected by excipients, environmental and processing stresses. We have used gabapentin, a structural analog of the neurotransmitter model compound to study the effect of manufacturing-related mechanical stress and environmental factors on solid state stability of crystalline drugs. Gabapentin undergoes intramolecular cyclization to form a gabapentin_lactam with accompanying loss of water. In addition, gabapentin is susceptible to physical transformations by mechanical and environmental processes involving the loss of crystallinity and formation of three polymorphs (Form II, III, and IV) and one monohydrate (Form I). In our studies, gabapentin was co-milled with various excipients in a planetary mill (Fritsch planetary micro mill Pulverisette 7) to induce crystal defects under mechanical stress. Then the gabapentin/excipient mixtures were stored at 50 ˚C in the presence of various relative humidities (5-50 %) to induce chemical cyclization. The rate of gabapentin_lactam formation was measured by HPLC. And the concentration time profiles were described by a degradation model that accounted for various physical and chemical transformations associated with lactamization, such as, autocatalytic branching, unimolecular dehydration and branching termination. The lactamization kinetics were dependent on excipients and moisture wherein excipients affected the extent of crystal defects during milling and the kinetics of lactamization upon storage. In general, exposure to high humidity decreased the rate of lactamization. Among the various excipients that were co-milled with gabapentin, silicon dioxide caused the most damage during mechanical stress A Kinetic Model for the Solid State Degradation of Gabapentin Z Zong, L Kirsch, A Kaushal, R Suryanarayanan, E Dempah, E Munson. National Institute for Pharmaceutical Technology and Education and The University of Iowa Purpose: Development of kinetic model to describe the solid-state degradation of gabapentin and the effects of manufacturing and environmental stress. Methods: Samples of crystalline gabapentin were milled in a planetary mill (Pulviserette 7, Planetary Micro Mill) for 0 to 240 minutes and stored at 25 to 60 °C and 5 to 90% RH). The time-dependent appearance profiles (measured by RP-HPLC) of the only degradation product (gabapentin-lactam, gaba-L) were used to developa quantitative kinetic model by non-linear regression. Polymorphic gabapentin transitions were examined by PXRD. III was observed when samples were milled for ≥ 120 minutes. Form III reverted to Form II, and the reversion was accelerated at high humidity. Kinetics of gaba-L formation upon storage at various temperatures and humidities were described by the following model: where G* represents fraction crystal-disordered substrate. Milling conditions affected the initial values of gaba-L and G* used to describe lactam formation kinetics. The estimated rate constant values were varied with temperature and humidity. Conclusion: Gaba-L formation occurred by a combination of first and second order kinetics. Milling stress was accounted for by the initial values of gaba-L and G*. Increased humidity dramatically increased the rate constant for branching termination whereas temperature affected all estimated rate constant values. Predicting Gabapentin Stability upon Processing using SSNMR K. E. Dempah1,2, A. Kaushal3, H. Huang4, R. Suryanarayanan3, L. Kirsch4, E. Munson2 1 University of Kansas, 2 University of Kentucky, 3 University of Minnesota, 4University of Iowa The objective of this study is to predict the stability of ball-milled gabapentin based on proton spin-lattice relaxation times and form conversions as measured by solid- state Nuclear Magnetic Resonance spectroscopy (SSNMR). As-received gabapentin (form II; anhydrous) was ball-milled for 45min in both the presence and absence of hydroxypropylcellulose (HPC). The samples were then stored for two days at 50°C either under 0% or 91% relative humidity and analyzed by 13C SSNMR and PXRD. HPLC was also run on the samples to determine the amount of degradation product formed before and after storage. The 1H T1 values measured for the sample varied from 130s for the as received unstressed material without HPC to 11s for the material that had been ball-milled in the presence of HPC. Samples with longer 1H T1 values were substantially more stable than samples that had shorter T1 values. Samples milled with HPC had detectable Form III (anhydrous) crystals as well. The exceptions were samples that were exposed to high humidity, which had fast relaxation times but slow degradation rates. The exposure to high humidity probably results in crystals repairing themselves to eliminate high-energy reactive sites. The result is smaller particles than the original material, but less reactivity because the healing process reduces the number of reactive sites. The exposure to high humidity also resulted in the Gabapentin (Form II) samples that were milled for 45min showed much shorter relaxation times and, in the presence of HPC, converted to Form III. These results suggest that SSNMR can be used to predict whether gabapentin will remain stable in a formulated product, as both relaxation times and form conversion can be determined within a formulation. The Effects of Source Variability and Lot-to-Lot Variability on the Dehydration of Trehalose Dihydrate Sarah J. Pyszczynski1,2 and Eric J. Munson2 1Department of Pharmaceutical Chemistry, University of Kansas, Lawrence, KS 66047 2Department of Pharmaceutical Sciences, University of Kentucky, Lexington, KY 40536 Purpose: To investigate the influence of source variability and lot-to-lot variability on the forms that are generated upon dehydration of trehalose dihydrate. Methods: Three lots of α,α-trehalose dihydrate were obtained from each of three manufacturers (nine samples total): Acros 18255, Fluka 90210, and Sigma T9449. Bulk materials were sieved to obtain particle size fractions of <75 µm, 75–125 µm, 125–180 µm, 180–425 µm, 425–850 µm, and >850 µm. Physical transformations of sieved fractions of trehalose dihydrate upon heating at 10 °C/minute were studied using differential scanning calorimetry (DSC), thermogravimetric analysis (TGA), and hot-stage microscopy (HSM). Bulk samples from three of the nine lots were dehydrated under vacuum at 75 °C, 100 °C, and 125 °C. Dehydrated samples were analyzed by DSC and 13C solid-state NMR spectroscopy (SSNMR). Results: Observations from DSC and HSM analyses suggest that the dehydration behavior of trehalose dihydrate upon heating can be classified in one of three ways: 1. Upon dehydration, β-anhydrate is generated in significant amounts for all particle sizes. Increasing amounts of amorphous material and α-anhydrate are formed as the particle size decreases. (One Sigma sample) 2. β-anhydrate is formed upon dehydration only for the larger particles. For smaller particles, amorphous trehalose and α-anhydrate are formed upon dehydration, and significant crystallization to the β-anhydrate occurs above 170 °C. (Fluka samples and two Sigma samples) 3. Similar to the behavior described in 2, but little to no crystallization to the β- Isothermal dehydrations were performed on one sample of each type. Different mixtures of anhydrous forms were generated from dehydrations of the same starting material at different temperatures as well as from dehydrations of different starting materials at the same temperature. With SSNMR, it was possible to detect forms that could not be detected with DSC. Conclusion: Both source variability and lot-to-lot variability significantly impact the forms that are generated upon dehydration of trehalose dihydrate. Regardless of the particle size, certain samples have a higher propensity to form the β-anhydrate, and other samples tend to form amorphous trehalose. Those that form amorphous trehalose display different tendencies to crystallize to the β-anhydrate. Variation in Product Temperature and Drying Time within a Lyophilized Batch during Primary Drying Benjamin J. Baranowski, Michael J. Pikal, Pooja Sane, Robin H. Bogner University of Connecticut, Storrs, CT It is current practice to develop a design space by defining the optimum operating conditions for primary drying and setting conservative operating ranges to assure “zero defects”. If, in fact, there are never any failures, the design space is probably set too conservatively. A six sigma strategy has 3.4 defects per million. When producing 10 billion units of a product over its lifetime, a six sigma strategy would predict 340,000 defective products. We aimed to determine the relationship between the design space for primary drying and the acceptable level of defects in a batch. An in silico model has been developed to propagate the variances in several operating conditions (i.e., shelf temperature and fill volume) along with known variances in product resistance and heat transfer coefficient through to the variation in the time to complete primary drying and the maximum product temperature during primary drying. As an example, the design space is defined here as the primary drying cycle time and mean shelf temperature that assure a certain level of dry product that does not exceed a temperature of -32.3°C. The area of the design space shrinks as 1) the acceptable level of defect is reduced and 2) the variances of input parameters increase. Since the variance in product resistance, in particular, is known to change during scale-up, adjustments in the design space are required to achieve the same level of defect. Hybrid Controls for a Fluid Bed Drying Unit Operation to Facilitate Full System
Automation and the Efficient Development of a Design Space
Brian M. Zacour, James K. Drennen III, and Carl A. Anderson
Duquesne University, Center for Pharmaceutical Technology, 600 Forbes Ave.,
Pittsburgh, PA 15282

Purpose: To develop a design space for the fluid bed drying unit operation to
demonstrate a portion of the QbD approach for pharmaceutical development using a
novel hybrid control system.
Methods: Experimental design techniques were used to develop a series of experimental
trials to study a combination of process factors using a Diosna Minilab. Final granule
characteristics and downstream process quality metrics were measured to develop models
to predict product quality from process parameters. Thermodynamic calculations were
used to control the internal process environment and also significantly reduce the number
of factors that need to be included in the experimental design. Near-infrared
spectroscopy (NIRS) is used to track the moisture content of the product throughout the
process and is used to define the endpoint of granulation and drying. Pressure drop
across the product bed is used to control the airflow velocity to maintain a constant bed
height. The process is fully automated, utilizing open process control (OPC)
communication, the control software SynTQ (Optimal), and the digital automation
system DeltaV (Emerson).
Results: The control system was shown to be successful in controlling the fluid bed
drying unit operation in an efficient manner. The thermodynamic calculations allowed
the internal processing environment to be controlled regardless of external environmental
fluctuations, while greatly reducing the dimensions of the experimental design. It was
shown that the thermodynamic environment is crucial in controlling the final granule
properties, and maintaining constant manufacturing parameters while ignoring external
fluctuations resulted in a variable drying environment, which affected the final granule
properties. Process models were created to predict the final tablet properties using the
experimental design input factors. This information was utilized to define a design space
to fit the needs of this formulation for this complex unit operation.
Conclusions: Robust process models were created using hybrid controls in an automated
system for a lab scale fluid bed drying process. The study is an example of a portion of a
design space integrated into the QbD approach for pharmaceutical unit operations
development.
Efficient Near Infrared Spectroscopic Calibrations for Pharmaceutical Blend Monitoring
Brian M. Zacour, Benoit Igne, James K. Drennen III, and Carl A. Anderson
Duquesne University, Center for Pharmaceutical Technology, 600 Forbes Ave.,
Pittsburgh, PA 15282
Purpose: To demonstrate the development of efficient near infrared (NIR) calibrations
using reduced sample sets for the initial model calculations of a pharmaceutical blend
monitoring system using several multivariate algorithms.
Methods: NIR calibration models were developed using only pure component samples,
pharmaceutical granules, and the final blend formulation using several multivariate
algorithms. These algorithms calculate the regression vector differently, and some are
more appropriate for use on reduced sample sets. The prediction statistics were
compared for each model using error estimates and blend monitoring of an independent
batch. Finally, the two models providing the best prediction statistics were transferred to
a different NIR instrument as an additional indicator of model robustness.
Results: A pharmaceutical analyst often has access to pure component samples for all
chemical constituents in a formulation. Consequently, multivariate algorithms can be
applied to a limited number of samples to develop NIR calibrations, which can
substantially lower the overall cost of development. This study demonstrates the utility
of an efficient NIR calibration for controlling the blending unit operation of an 8-
component pharmaceutical formulation. A robust model was created using a reduced
sample set processed on a small scale blend simulator. As an analyst encounters further
samples, the original calibration model can be easily updated to expand the variance
space as necessary.
Conclusions: This study demonstrates the ability to create sensitive and robust
multivariate spectral calibrations using limited sample sets when pure component scans
are available. Classical least-squares based methods produced the most sensitive results
in these situations due to their ability to quickly relate the regression vector to the net
analyte signal.

Source: http://www.nipte.org/docs/NIPTE%20Research%20Conference%20Abstracts.pdf