MedicineLakex

 

Lpdt11(04).book(lpdt_a_176964.fm)

Pharmaceutical Development and Technology, 11:403–408, 2006 Copyright Informa HealthcareISSN: 1083-7450 print / 1097-9867 onlineDOI: 10.1080/10837450600770072 Assessment of Dynamic Image Analysis as a Surrogate Dissolution Test
for a Coated Multiparticulate Product
Dynamic Image Analysis Grant Heinicke
Formulation Development, Actavis, Elizabeth, NJ, USA
Joseph B. Schwartz
Department of Pharmaceutical Sciences, University of the Sciences in Philadelphia, Philadelphia, PA, USA
in drug release with changes in particle size distribution Dynamic image analysis (DIA) was used to measure parti- (PSD) on a model system.
cle diameter (D50) of in—process samples removed during fluidbed coating. A single, rapid measurement gave D50 to within4 μm. Samples removed at intervals of 2% weight gain werereadily distinguishable by DIA and by their drug release profiles.
MATERIALS AND METHODS
Drug release was related to D50. DIA was assessed as a surro-gate dissolution test with considerable potential. Current limita- A description of the model system used for this work tions of the approach were presented.
was reported previously.[4] Briefly, the system consists ofdiltiazem hydrochloride (Profarmaco, Milan, Italy) layered dynamic image analysis, drug release, Wurster onto sugar spheres (30–35 mesh, Paulaur, Cranbury, NJ) coating, multi-particulate, pellet using a solution of HPC (Hercules, Wilmington, DE) inalcohol (Equistar, Houston, TX) as a binder. The polymercoat consists of Eudragits RS and RL (Degussa, Piscataway,NJ), triethyl citrate (Moreflex, Greensboro, NC), sodium lauryl sulfate (Stepan, Northfield, IL), and talc (Luzenac,Denver, CO). The coat was applied from alcohol. Cores Fluid bed coated multiparticulates constitute an were made in a GRG-100 rotor in a GPCG-200 fluid bed, important class of solid oral dosage forms. Preparation of and pellets were made in a 4″ Wurster in a GPCG-3 fluid these dosage forms has been well described in the litera- bed (GlattAir Techniques, Ramsey, NJ) using the condi- ture.[1,2] The particles of interest in this article are reservoir tions given in Table 1.
devices, consisting of rapidly dissolving cores that are Seven repeat coating experiments were conducted subsequently coated with a polymer membrane to control from a single lot of diltiazem cores. Two bulk coating drug release. As controlled-release coating material is solutions were prepared; 1244 g of coating solution was applied in a fluid bed, polymer membrane thickness dispensed from either bulk for each coating experiment increases,[3,4] resulting in a decrease in drug release as taking care to ensure that dissolved solids were equal at measured by dissolution testing.[5–9] Changes in mem- the time spraying by replacing solvent lost to evaporation.
brane thickness can be measured by dynamic image analy- Samples were removed at 7, 9, 11, 13, 15, 17, and 19% sis (DIA).[10] In this work we demonstrate the potential of theoretical polymer coat weight (TPCW) via a sample port DIA as a surrogate dissolution test by correlating changes without interruption to the spraying process. Multipleinsertions of the probe were used to remove between 11and 14 g for DIA and dissolution analysis. TPCW was cal- Received 21 December 2005, Accepted 3 April 2006.
culated as weight of dissolved coating solids applied over Address correspondence to Grant Heinicke, Formulation core weight plus dissolved solids weight × 100%. No cor- Development, Actavis, 200 Elmora Avenue, Elizabeth, NJ 07207; rections were made for the weight of samples removed.
Tel.: 908-659-2599; Fax: 908-659-2390; E-mail: [email protected] Coating efficiency was calculated as measured weight G. Heinicke and J.B. Schwartz
Table 1
the replicates for the time to 50% released was generally Manufacturing conditions less than 4 min, with a maximum difference of 10 min. Linearregressions were determined by using Microsoft Excel 2000 or JMP software, version 5.1.1 (SAS Institute Inc., Cary, NC). Statistical calculations were performed by using JMP.
Process air volume RESULTS AND DISCUSSION
Coating efficiency ranged between 92.7 and 98.2% for the seven coating experiments. Measured weight gains on small experiments are susceptible to loss of particles during sampling and discharge, and the range of coating efficiencies was not considered excessive. D50 was determined on eachTPCW sample from each coat application (49 samples).
Results are presented in Table 2. Each PSD result comes gain over theoretical weight gain × 100%. DIA was con- from approximately 15,000 particles, and measurement was ducted on a Camsizer (Horiba Instruments, Irvine, CA), complete within 3–4 min of removal of the sample from the and the median value (D50) was taken from the PSD process. A single PSD measurement was used for each sam- report generated. A single PSD measurement of each sam- ple to demonstrate the real–time capability of the instrument ple was used except where noted in the text. The Camsizer with the intention of in-process use of the results.
was calibrated before use by using the vendor supplied The variation in D50s at any TPCW for the seven calibration reticule (ISO9000 traceable standard). Dissolu- experiments (columns of Table 2) spanned a maximum of tions were performed in USP Apparatus II (Hanson SR6, 4 μm. This was the extent of variation on repeat measure- Chatsworth, CA, or Distek 5100, North Brunswick, NJ) at ments of a single sample (Figure 1, footnote to Table 2) 100 rpm using 900 mL of helium degassed 0.1 N HCl and from other work.[4] Repeat measures of a single sample equilibrated to 37°C. A quantity of pellets containing 300 mg resulted in a standard deviation of about 1 μm (Table 2), of diltiazem HCl was added to each dissolution kettle.
but increased the time required to obtain a D50 value. For Media were withdrawn at predetermined times automati- in-process use, a compromise must be struck between the cally, and diltiazem concentration was measured by UV at time taken to get a result and progress of the batch.
238 nm. The time to release 50% (T50) was interpolated Typical PSD data for a series of samples from a single from the line of best fit through the linear portion of the coating experiment are shown in Figure 2.[11] The 2% drug release profile. Each reported drug release profile is TPCW increments from the coating experiments resulted the average of at least two replicates. Agreement between in uniform translations of PSDs along the size axis. This Table 2
D50 (μm) results for seven experiments at each of seven coat weights Experiment no.
aEight measurements of this sample were 994, 994, 995, 995, 994, 996, 994, and 998, average of 995.0 ± 1.41 μm. This apparent size difference was attributed to the absence of dynamic calibration and the time lapse between the sequences of measurements,giving ample opportunity for accidental alteration of the guide (see text). NA, not applicable.
Dynamic Image Analysis
Because the Camsizer is nondestructive, each of the samples measured in Figure 2 was recovered and tested for drug release. The data are presented in Figure 3.[11] The 2% TPCW increase of each successive sample increased T50 by about 1 hr. This relationship is shown in graphically in Figure 4.
The linearity of the D50 changes with TPCW in Table 2 combined with the linearity of T50 changes with TPCW in 900 1000 1100 1200 Figure 4 results in a linear relationship between T50 and Size Classes (um)
D50. The relationship between T50 and D50 for experi- Repeat PSD measures of the 15% TPCW sample ment 3 is shown in Figure 5. The line can be used to calcu- from Experiment #3.
late T50, with units of hours, from D50 measurements thatare accessible in minutes.
This plot can be used as a standard curve for subse- quent samples. From the slope of the line in Figure 5, each micrometer change in D50 resulted in about 13-min change in T50. The D50s of the seven samples with 15% TPCW ranged from 998 to 1002 um (Table 2). Drug release from the 15% TPCW samples was measured (Figure 6), and T50s ranged from 528 to 557 min. The reproducibility of manufacture, of drug release, and of D50 measurements shown in these examples supports 900 1000 1100 1200 the use of D50 measurements for calculation of T50. In Size Classes (um)
PSDs from which D50 data were taken for Experiment #3.
sequence has been observed for all Camsizer measure- ments on coating of rotor–granulated cores conducted in our laboratory, leading to confidence that random error is not confounding data interpretation within a series of mea- surements. Based on the work of Paine[12] and Parrott,[13]the numbers of particles in the DIA measurements in this work were sufficient to adequately represent the batch, Time (min)
particularly because PSDs of drug layered on sugarspheres are normal[10,14] and reasonably narrow. In addi- Drug release profiles from all TPCW samples from tion, fluid bed processes and in-process sampling have Experiment #3.
been shown to be sufficiently random to allow samplingfrom the sample probe in the manner described in this y = 58.054x - 317.7 Uniformity of translation of the PSD histograms with TPCW in Figure 2 and the other six experiments can be appreciated by plotting the data in Table 2. The plots had the same slopes (Appendix 1) with r2 values between 0.9945 and 0.9992. The linearity of these relationships, as opposed to the theoretical cubic relationship of adding volume to a sphere, was explained by the small size increase due to coating relative to the starting diameter of the particles themselves. Link and Schlunder[15] made a Coat Weight (%)
similar approximation for a coating process without signif-icant error.
T50 versus TPCW for experiment #3.
G. Heinicke and J.B. Schwartz
rapid rate of diameter change is not anticipated with larger batch sizes. A typical commercial scale process would y = 12.864x - 12326 apply around 0.2% TPCW in 5 min.
The calibration method recommended by the vendor consists of taking an image of a reticule in a fixed plane at optimum distance from the screen. The computer then self-adjusts for the known image sizes. Particle size measurement itself is a dynamic process in which parti- cles flow with a distribution of trajectories with respect to screen distance. Consequently, image size is a func- tion of particle trajectory during measurement but not during calibration. Adding a guide narrows the distribu- T50 versus D50 for Experiment #3.
tion of particle trajectories during measurement, but theposition of the guide itself translates the PSD along thesize axis. Although measurement of a series of sampleswas accurate if made sequentially, there is no assurance other words, DIA was a suitable surrogate for dissolution that the position of the guide did not eventually get altered. An alternative calibration method requiring a The same lot of cores was used throughout this work dynamic measurement of standard particles would over- to assess the reproducibility of the processes of coating, come this limitation, and this improvement was dis- dissolution, and image analysis without the influence of cussed with the vendor. Such a calibration would ensure core lot–to–lot variability. Reducing variability was pre- conformity over time to a standard curve such as that in ferred for initial assessment of DIA as a surrogate for dis- solution although it was understood that materials,[16]processing equipment and processing conditions[17,18]determine size and surface morphology of cores,[19–23]which in turn affect drug release. Polymer coat function is known to be sensitive to coat thickness[24] and processingparameters[17,25,26] and can even be sensitive to seasonal DIA was used to measure D50 of a multiparticulate variation[27] if adequate controls are not in place. Use of sample of 15,000 individuals within minutes of being DIA as a surrogate for dissolution requires that all these removed from a Wurster coating process. PSD measure- influences be under control. In addition, although the ments on samples containing this number of particles are D50 measurement was available within 4 min of removal expected to adequately represent the population from of the sample from the process, 2% TPCW was applied which they are taken. D50 results from single measure- in approximately 11 min in these experiments. Therefore, ments were used to generate a relationship between size around 1% TPCW was applied as D50 was measured, and the drug release metric of T50. The relationship was resulting in an increase of T50 by about 30 min. Such a linear. It was shown that D50 results, accessible withinminutes, could be used to predict drug release (T50)results with considerable savings in testing time and effort.
If surrogate dissolution capability by DIA is sought, for- mulations may be designed so that drug release is lesssensitive to coat thickness than is this system. Control of material, formulation, and process variables that affect coat function with respect to thickness are required for application of this method.
This work was supported by Alpharma/Purepac. The Time (min)
authors thank Dr. Stan Deming of Statistical Designs, Drug release from the 15% TPCW sample from each Houston, TX, and Dr. Garth Boehm of Actavis for many Dynamic Image Analysis
19. Vertommen, J.; Rombaut, P.; Kinget, R. Internal and exter- nal structure of pellets made in a rotary processor. Int. J.
1. McGinity, J. (Ed.). Aqueous Polymeric Coatings for Phar- Pharm. 1998, 161, 225–236.
maceutical Dosage Forms, 2nd Ed., Revised and Expanded; 20. Rashid, H.A.; Heinamaki, J.; Antikainen, O.; Yliruusi, J.
Marcel Dekker: New York, 1997, 81–100.
Influence of the centrifugal granulating process on the prop- 2. Ghebre-Sellassie, I. (Ed.). Multiparticulate Oral Drug Deliv- erties of layered pellets. Eur. J. Pharm. Biopharm. 2001, 51,
ery; Marcel Dekker: New York, 1994, 405–456.
3. Kramar, A.; Turk, S.; Vrecer, F. Statistical optimization of 21. Sinchaipanid, N.; Chitropas, P.; Mitrevej, A. Influences of diclofenac sustained release pellets coated with polymethacrylic layering on theophylline pellet characteristics. Pharm. Dev.
films. Int. J. Pharm. 2003, 256, 43–52.
Technol. 2004, 9 (2), 163–170.
4. Heinicke, G.; Schwartz, J.B. Particle size distributions 22. Lorck, C.A.; Gruenberg, P.C.; Junger, H.; Laicher, A. Influ- of inert spheres and pelletized pharmaceutical products ence of process parameters on sustained-release theophylline by image analysis. Pharm. Dev. Tech. 2004, 9 (4),
pellets coated with aqueous polymer dispersions and organic solvent-based polymer solutions. Eur. J. Pharm. Biopharm.
5. Sadeghi, F.; Ford, J.L.; Rubinstein, M.H.; Rajabi-Siahboomi, 1997, 43, 149–157.
A.R. Study of drug release from pellets coated with Sure- 23. Sousa, J.J.; Sousa, A.; Moura, M.J.; Podczeck, F.; Newton, lease containing hydroxypropylmethylcellulose. Drug Dev.
J.M. The influence of core materials and film coating on the Ind. Pharm. 2001, 27 (5), 419–430.
drug release from coated pellets. Int. J. Pharm. 2002, 233,
6. Mehta, A.M. Factors in the development of oral controlled-release dosage forms. Pharm. Manuf. 1986,
24. Zhang, G.; Schwartz, J.B.; Schnaare, R.L. Bead coating. I.
January, 23–29.
Change in release kinetics (and mechanism) due to coating 7. Frenning, G.; Tunon, A.; Alderborn, G. Modelling of drug levels. Pharm. Res. 1991, 8 (3), 331–335.
release from coated granular pellets. J. Control. Release 25. Yang, S.T.; Savage, G.V.; Weiss, J.; Ghebre–Sellassie, I.
2003, 92, 113–123.
The effect of spray mode and chamber geometry of fluid– 8. Rekhi, G.S.; Porter, S.C.; Jambhekar, S.S. Factors affecting bed coating equipment and other parameters on an aque- the release of propranolol hydrochloride from beads coated ous–based ethylcellulose coating. Int. J. Pharm. 1992, 86,
with aqueous polymeric dispersions. Drug Dev. Ind. Pharm.
1995, 21 (6), 709–729.
26. Yang, S.T.; Ghebre–Sellassie, I. The effect of product bed 9. Maejima, T.; McGinity, J.W. Influence of film additives on temperature on the microstructure of Aquacoat–based con- stabilizing drug release rates form pellets coated with acrylic trolled–release coatings. Int. J. Pharm. 1990, 60, 109–124.
polymers. Pharm. Dev. Tech. 2001, 6 (2), 211–221.
27. Jones, D. Air suspension coating for multiparticulates. Drug 10. Heinicke, G.; Matthews, F.; Schwartz, J.B. The effects of sub- Dev. Ind. Pharm. 1994, 20 (20), 3175–3206.
strate size, surface area, and density on coat thickness of
multi-particulate dosage forms. Pharm. Dev. Tech. 2005, 10
(1), 85–96.
11. These data were presented in a poster at AAPS in Baltimore, APPENDIX 1
12. Paine, A.J. Error estimates in the sampling from particle Testing the significance of the slopes was done by the size distributions. Part. Part. Syst. Charact. 1993, 10,
method of Extra Sums of Squares. A model was fitted to the entire data set relating TPCW and D50. Two additional mod- 13. Parrott, E.L. Milling. In The Theory and Practice of Indus- els were fitted to the data, each successively adding a term for trial Pharmacy, 3rd Ed.; Lachman L.; Lieberman, H.; Kanig,
J. Eds.; Lea & Febiger: Philadelphia, 1986, 21–46.
14. Borgquist, P.; Zackrisson, G.; Nilsson, B.; Axelsson, A.
Simulation and parametric study of a film–coated controlled–release pharmaceutical. J. Control. Release 2002,
80, 229–245.
SS 15912.020 Error 15. Link, K.C; Schlunder, E. Fluidized bed spray granulation.
Investigation of the coating process on a single sphere.
Chem. Eng. Process. 1997, 36, 443–457.
16. Armstrong, N.A. Functionality related tests for excipients.
Int. J. Pharm. 1997, 155, 1–5.
17. Jones, D.M. Factors to consider in fluid-bed processing.
Pharm. Technol. 1985, April, 50–62.
18. Guignon, B.; Regalado, E.; Duquenoy, A.; Dumoulin, E.
Helping to choose operating parameters for a coating fluid Sum of squares tree for the Reduced Model, Expanded bed process. Powder Technol. 2003, 130, 193–198.
Model, and Further Expanded Model of the data in Table 2.
G. Heinicke and J.B. Schwartz
intercepts for individual experiments and then slopes for indi- is the sum of squares from the model, vidual experiments. The models are termed "the Reduced is the error unexplained by the model, and Model," "the Expanded Model," and "the Further Expanded is the degrees of freedom for the terms in the Model." A "Sum of Squares Tree" was constructed (Figure numerator. Significance of the F values was obtained from A1) and F values were calculated by using Eq. (1) tables. For example, F for the Expanded Model is (24.408/6)/(68.837/41) = 2.42, which is significant for 6 and 41degrees of freedom at the 95% level. F for the further expanded model was 1.26, which was not significant for 6 and 35 degrees of freedom at the 95% level. It was con- cluded that the slopes of the lines were not statistically sig-

Source: http://www.christison.co.uk/cms/thesite/public/uploads/tblproducts_file/1183984809_512.pdf