Scaling of Fluid Bed Coating

Scaling of Fluid Bed Coating a report by Harlan S Hall Coating Place, Inc. The Wurster Process (fluid bed coating) is widely used and has become a primary method for the microencapsulating of particles. The original patents were filed in 1953 1 and, over time, the process has evolved into one of the most adaptable means of coating small particles. Although the process has been widely used for more than 30 years, some misconceptions regarding scaling this process from the laboratory to commercial production still exist. One published example 2 shows little correlation between load size, spray rate and process time, with total spraying time increasing at every scale over five different size chambers, both single and multiple nozzles. Other published studies, which confirm reproducibility of coating applied, 3,4 also demonstrate a processing time increase of 5x and 3.1x for two different products scaled from small scale to manufacturing scale (3kg to180kg). In working with companies who need to scale up newly developed products from laboratory or research to manufacturing scale, we have been made aware of the difficulties encountered to make the proper changes to accomplish the task. We have had clients who believe that linear scale-up of the fluid bed process is not possible. In one case, after demonstrating successful linear scalability of the process, the client insisted on additional batches of product to prove that the initial work was not a fluke. We recognise that knowledgeable and skilled people have scaled up a variety of products successfully. It is also clear that there remains some mystery about the process, most likely based on bad experiences. The purpose of this article is to remove the mystery behind the scale-up of fluid bed processes. In various discussions regarding the scale-up of fluid bed coating systems, the statement has been made, correctly, that the new spray rate can be calculated by multiplying the old spray rate times the ratio of new airflow to old. 3,5 Formula 1: R 2 = R 1 x (V 2 /V 1 ) where R = spray rate V = air volume In 2002, an equipment company published information 6 stating, again correctly, that linear air velocity in the equipment should be kept approximately constant as the process is scaled up. Because the base plate area increases as the process is scaled up, keeping the linear velocity constant also means that the air volume will increase in proportion to the base plate area. Formula 2: 2a: V = L x A L = (V/A) L = linear flow rate in ft/min V = air volume in ft 3 /min A = base plate area in ft 2 Combining these statements tells us that, at least in theory, the scaled-up spray rate should be proportional to both the increased plate area and the increased air volume. In fact, data 2,4,6 8 and many anecdotal experiences indicate that it is not that simple. Formula 3: R 2 = R 1 x (A 2 /A 1 ) = R 1 x (V 2 /V 1 ) There are several reasons why Formula 1 and 2 are correct, but insufficient evidence exists to assure us 1. D Wurster (1953), Method for Applying Coating to Tablets, US Patent 2,648,609. 2. Atul Mehta, Scale Up Considerations in the Fluid Bed Process for Controlled Release Products, Pharmaceutical Technology, Feb. 1988. 3. Kinsey, et al., (2002), Use of Experimental Design in the Scale Up of an Aqueous Ethylcellulose-based Coating Formulation, Colorcon, website (http://www.colorcon.com). 4. Colorcon website (http://www.colorcon.com), Application of SuRelease to Chlorpheniramine Maleate Drug-Loaded Pellets: Scale Up, 2002. 5. Stuart Porter, Scale Up of Extended Release Dosage Forms, AAPS Workshop, Sept., 1992. 6. Scale Up Factors in Fluid Bed Processing, Fluid Air Inc. website (http://www.fluidairinc.com), 2002. 7. Wisconsin Alumni Research Foundation, Internal Document, 1969. 8. D Jones (1994), Air Suspension Coating, International Society of Pharmaceutical Engineering. 1

Figure 1a: Maximum Droplet Size: Test Nozzle A Figure 1b: Maximum Droplet Size: Test Nozzle A that Formula 3 is completely correct and leads to satisfactory scale-up. It is recognised that scaling up from a one-litre chamber size to a 90-litre chamber size using a single nozzle is fairly easy and predictable. The maximum distance for efficient delivery of atomised spray to the surface of particles is approximately 4.7 (12cm) radius or 9.4 (24cm) diameter, approximately equivalent of a 9 (23cm) centre partition inside an 18 (46cm) coating chamber). 2,5 7,14,15 A single nozzle distributes the coating material efficiently to the particles, as evidenced by the proportional relationship between plate area, air volume and spray rate observed routinely 5,8,9 up to a coating chamber diameter of 18 to 19 (46cm to 49cm). The primary factors over this range are nozzle function and plate design. As process scale is increased using a single nozzle, it is essential that the nozzle used is capable of atomising the coating formulation efficiently, even as the rate of liquid delivery increases. At the smaller scale, liquid flow rates are modest, perhaps 5g/min to 20g/min, and many commercially available nozzles work well. As the scale and flow rates increase, the nozzle must still atomise effectively. At higher flow rates (300g/min to 800g/min), only a few well-designed nozzles are capable of delivering a uniformly atomised mist of small droplet size (10µ to 30µ). If the nozzle you are working with is incapable of meeting this performance standard, the scale-up will fail, even if fluidising air volume is proportioned correctly. Large droplets created by lower-performance nozzles are not distributed as evenly and do not dry as quickly as smaller droplets. In these cases, the nozzle itself becomes rate-limiting. 9 12 When the spray rate exceeds the capacity of the nozzle to maintain uniform atomisation, large droplets (up to 80µ) appear along with the finely atomised droplets (compare median with maximum droplet size). The presence of large droplets results in the formation of agglomerates. A comparison of the graphs reveals, 12 for example, that a nozzle delivering 500gr/min of coating has a median droplet size of 15µ (80psi) and will include droplets as large as 35µ. Reducing the spray rate to 300gr/min with the same nozzle will reduce the median only to 13µ, but will reduce the maximum to 26µ. The significance of this reduction is demonstrated by Figure 2 comparing the growth of particles by agglomeration as a function of spray rate. It was stated previously that scaling up with a single nozzle to approximately 18 (46cm) chamber diameter is relatively easy and straightforward. Once the scale is increased beyond an 18 (46cm) diameter, the ability to deliver the spray to all the particles drops rapidly as distance increases. 10,13 15 This is likely due to the dense flow of particles around the nozzle, which effectively intercepts all the spray droplets. Anything circulating farther from the nozzle than the last intercepted droplet does not receive any spray and thus is effectively not in the process (for that particular cycle). Increasing chamber diameter beyond this point does result in larger batch sizes, but does not result in faster throughput, even when the larger plate area and air volume are taken into consideration. 2 9. Coating Place, Inc., Nozzle Study, Internal Study, October 1974. 10. Coating Place, Inc., Small Particle Coating, Internal Study,1977. 11. Lord & Hall, Uniformity of Coating on Small Particles, Interphex USA, 1988. 12. Coating Place, Inc., T Breunig, Atomized Droplet Size Distribution, Internal Study, 2003. 13. D Jones (1993), Fluidized Bed With Spray Nozzle Shielding, US Patent 5,236,503. 14. D Jones (1995), Fluidized Bed With Spray Nozzle Shielding, US Patent 5,437,889. 15. A Avhandling, On the Optimization of the Fluid Bed Particulate Coating Process, PhD Thesis, Dept of Chem. Engineering & Technology, Royal Institute of Technology, Stockholm, Sweden, 1998.

Scaling of Fluid Bed Coating In an attempt to minimise agglomeration, some configurations 13,14 prevent particles from approaching the spray orifice closely by creating a void area around the nozzle. This void is typically created by a baffle or shield inserted around the nozzle itself. This shield also allows the use of faster spray rates. It effectively increases the length of time that droplets remain airborne, permitting them to dry partially and/or completely before contacting the surface of the particles to be coated. The end results are spraydried coating materials and particles that have a more porous film as measured by permeability of the film deposited. Figure 3 16 displays the difference in barrier properties for a product produced using a conventional Wurster (slower release) and utilising a shield around the nozzle (faster release). Efforts to offset this effect by diluting the solution with more solvent are counter-productive to faster processing. Figure 2: Particle Size Growth (Agglomeration) as Function of Spray Rate and Coating Level Figure 3: % Core Released vs. Time; Code NTNC Both curves are averages of multiple batches prepared using the same core material, coating level and precisely the same coating composition. In 1971 and 1972, Wisconsin Alumni Research Foundation (WARF) Coating Laboratory built a prototype multiple nozzle unit. The base module for this unit was a single 18 (46cm) chamber the largest effective size for a single nozzle unit. This unit had seven spray zones, based on a hexagonal pattern of six spray zones plus one additional zone placed in the centre. 18 As illustrated in Figure 4a, if the modules are simply placed into this pattern, there are large areas (shaded) that are outside the module diameters. 19 If no changes are made to this pattern, the shaded area contributes greatly to plate area (and air volume), but is not effectively part of the coating module (spray rate). This results in a mismatch between air volume and spray rate proportionality. Formula 4: R 2 R 1 x (A 2 /A 1 ) R 1 x (V 2 /V 1 ) The size and importance of the shaded area can be reduced by crowding the working zones together so that they overlap; however, the overlapping areas reduce the total area within the working zones (see Figure 4b). It is necessary to adjust the degree of overlap and to modify the plate design to balance these effects and restore proportionality. This is true for any number of nozzles and spray zones greater than one. If the chamber and plate design have been adjusted and balanced correctly, then each working zone is equivalent to a single nozzle chamber approximately 18 (46cm) in diameter, and total chamber volume is approximately a multiple of a single chamber. Since every working zone contains a single nozzle, each nozzle must only perform as it does in a single nozzle unit no scale-up of nozzle performance beyond the largest single nozzle unit is necessary. In the case of seven nozzles, as shown in Figures 4a and 4b, the optimised plate area and air volumes are seven times that of a single nozzle base unit. The seven nozzles each deliver and atomise the same volume of liquid as in the base module for a composite rate that is also seven times. This proportionality holds for any number (n) of coating zones. With the chamber and plate design established, it remains necessary to make some further adaptations to the equipment. In a single nozzle unit, there is only one working zone and all of the process air used must pass through it. When there are multiple working zones (any number n >1), the process air 16. Coating Place, Inc., Internal Data, Comparative Release Data. 17. Mehta and Jones, Coated Pellets Under the Microscope, Pharmaceutical Technology, 9(6), 1985, pp. 52 80. 18. Wisconsin Alumni Research Foundation (Coating Place, Inc.), Coating Uniformity of 46 Unit, 1973. 19. Hall, Scale Up: Wurster Fluid Bed Coating, Controlled Release Society, Annual Meeting, Seoul, Korea, 2002. 3

Figure 4a: Module Packing Patterns, 7x Figure 4b: Module Packing with Overlap, 7x Table 1: Nutritional Product Improved Shelf Stability (Code TTAF) Unit 4/6 Factor 18 Factor 46 Load, Kg 1 35x 35 7.1x 250 Fluidising Air (scfm) 19 31x 600 7.7x 4,600 Pump Rate g/min 8.8 29x 255 7.1x 1,800 (257/noz) Nozzles # @ psi 1 @ 15 1 @ 80 7 @ 80 collapse due to insufficient airflow. It is critical to prevent this from occurring. Introduction of a balancing mechanism in the plenum below the base plate limits the quantity of air that may reach any one zone. This not only prevents material being blown out of one zone, but effectively balances the airflow across the bed and positively enforced airflow throughout all zones. Table 2: Pharmaceutical Taste Masking, Site Transfer and Scale-up (Code TCIE) Unit 18 Factor 32 Load, Kg 28 3.2x 90 Fluidising Air (scfm) 730 3x 2,175 Pump Rate Grams/min 375 3x 1,125 Nozzles # @ psi 1 @ 80 3 @ 80 Table 3: Client Laboratory to Production Scale (Code TCIF) Client CPI CPI CPI Original Partial Full 3 Nozzle Unit 9 18 18 32 Load, Kg 3 5 20 28 90 Coat Time 210 Minutes 280* 195* 273* 275* is distributed among the zones and must be balanced and positively maintained. Air (a fluid) will tend to take the path of least resistance. If material is blown out of one zone, effectively removing the resistance of the particle load, air will begin to escape through that zone, resulting in lower airflow in the remaining zones. Should this occur, one or more of the remaining zones may When air volume and spray rate are proportional, and base plate, chamber design and air balance are designed correctly and under control, scale-up from single to multiple nozzle units is quite simple. Load size, plate area, air volume, number of nozzles and total spray rate are all a direct multiple of the parameters developed in a single nozzle unit. Such a unit is equivalent to (n) single nozzle coating units operating side by side. It is possible to alter batch size (bed depth) intentionally. The velocity of air required to suspend a particle is primarily a function of particle size, 20 not the depth of the bed. Unless particle size changes significantly, the linear air velocity required to fluidise the particles is not a function of bed depth, thus total air volume should not change. Because particle diameter increases as the cube root of mass (vol sphere = (4/3) r 3 ), particle size is not expected to change markedly for most coating applications. If air volume is constant, the spray rate should also be constant. Total mass fluidised will increase with bed depth, and more energy (blower rpm or % flap opening) will be required in order to maintain the intended air volume (velocity) as the mass increases. Having a greater load volume increases the quantity of coating required (total quantity required, not the coating level). If the quantity of coating required increases and spray rate remains the same, it would be reasonable to expect the process time to increase, even though scale-up parameters are linear. The author has 4 20. Handbook of Chemistry & Physics, Vol. 67, CRC Press, p. F-231.

Scaling of Fluid Bed Coating observed operators who attempted to increase spray rate when increasing load size, usually with adverse results. The effective maximum load size is limited by bowl (chamber) size and by partition height when partitions are used. Tables 1 3 show data from several actual products scaled up successfully in equipment with properly designed plates and chambers. Specific products are not identified. To preserve confidentiality, each is identified only by a four-letter code. Actual load sizes were adjusted for convenience in weighing (Table 1, large load rounded from 245kg to 250kg). when utilising properly designed base plates and chambers and when the base plate area/fluidising air volume/spray rate relationships are kept in proportion. If these factors are controlled, there is little need to make disproportionate changes in the basic operating conditions established at the smaller scale. Many reported difficulties in scaling up a process can be traced to improper correlation of these factors and/or to poor equipment design. Contact Information In Table 3, during experimental laboratory-scale trials, load size and coating level were varied, resulting in a range of process times. In the final manufacturing process, times are within the original ranges tested at small scale. Scale-up is linear even as load size varies. Conclusion Linear scale-up of fluid bed coating processes over a wide range of equipment size is readily achievable Frederick A Schulze Director, Sales and Marketing, Coating Place Inc. 200 Paoli Street Verona, WI 53593 Tel.: +1 817 477 2766 Fax: +1 817 453 1696 e-mail: fschulze@encap.com http://www.encap.com 5