A common objective in pharmaceutical processing is the removal of solids from fluid suspensions through filtration. The usual purpose is the removal of the solid particles to a specified extent, within a given time interval, at the largest possible throughput. Attainment of those goals is managed by proper selection of filtration conditions: principally an adequate effective filtration area (EFA) as defined by filter porosity and a proper rate of flow as regulated by applied differential pressure (ΔP) over the period of filtration. Were the fluid “clean,” by definition free of particles whether of microbial or of other origin, the task would be amenable to mathematical analysis. Flow rate would be directly related to ΔP over time and also to the EFA’s porosity in terms of pore numbers, dimensions, and so on. Time required could be calculated from the batch size processed over the duration.
PRODUCT FOCUS: BIOLOGICS
PROCESS FOCUS: DOWNSTREAM PROCESSING
WHO SHOULD READ: MANUFACTURING, ANALYTICAL, AND PROCESS DEVELOPMENT
KEYWORDS: MEMBRANE FILTRATION, MEMBRANE CHROMATOGRAPHY, FLOW RATE, ADSORPTIVE CAPTURE, PREFILTRATION
But when commonly processed liquids—those bearing suspended particles—are involved, the blocking and clogging of filter pores by retained particles change the equation. The pore size rating of a filter should be selected to retain the objectionable particles by sieving, and the aptitude of its polymeric composition for adsorptive sequestration of those particulates also needs to be known. The quantity and nature of retained particles requires accommodation in filtrative removal if the outcome is to be considered successful. Too extensive a particle load will prematurely block a filter’s delivery of sufficient throughput to meet the filtration’s goal. This equates with enough drug product to provide a adequate monetary return. Drug processing thus represents a technoeconomic challenge.
As filtration of a liquid progresses, suspended particles (whether organisms or otherwise) are arrested by sieving at the filter pores or adsorption to the membrane or nonwoven structure of the filter. Contrary to earlier teachings, membranes are not limited to surface retention because of their thinness (~120–150 µm). Two outcomes await particles small enough to enter the pores or fiber matrix: They may be conveyed by the suspending liquid close enough to the matrix wall and become adsorptively fixed to it, or they may escape such capture to emerge in the filtrate.
When particles are arrested because of their size or shape, leading to near complete pore blockage, the effect on fluid flow is immediate. When particles gather at the walls (eventually to clog the passageway), the flow is diminished more slowly. Either way, it is particle capture that interferes with and ultimately terminates a filter’s service life and defines its throughput. The term filter cake is usually reserved for a particle mass located on the filter surface. Plotting the rate of flow diminution over time can help ascertain mathematically a filter’s particle retention mechanism (1,2).
A complete cessation of flow is not sought. When the initial flow rate has decreased by about 80%, a point of diminishing return is usually considered to have been reached. It is considered impractical in terms of labor and costs to go beyond that point. In some applications, the filter can be used only to 50% blockage because progress beyond that point can cause yield losses due to bridging and unwanted sieve or adsorptive separation.
In the event of particulate loading in excess of a filter area’s ability to accommodate, fluid flow will be blocked prematurely, and a batch filtration will be incomplete. The consumed filter must be replaced before filtration can continue. Such an action, arduous in itself, is best prevented when product sterility is required. It could expose an entire filtration train to the needless risk of asepsis. So the EFA necessary to process an entire batch is calculated from flow-decline studies before processing activities are initiated (3).
Such studies use frequent sampling and filtrate analysis and can also indicate what blockage rate can be allowed without yield loss. However, flow-decline studies are too often based on models using flat filter disks 47 mm in diameter. They are seldom adequate prototypes for pleated cartridges. Carefully designed and implemented flow-decline studies are needed to secure dependable results. Studies with 47-mm discs are commonly recognized as “indicator trials” that require follow-up “verification trials” using small-scale pleated devices to establish a sound foundation for scaling work (3).
Interactions of EFA, ΔP, and Time
A filtration operation involves an equivalence of effects among the EFA, differential pressure, and processing time. This interdependency offers alternatives to achieving filtration goals: namely, an expeditious rate of filtration leading to as generous a throughput as possible sufficiently free of objectionable particles. For clean liquids when one of the three factors is maintained constant, increasing another necessitates a corresponding decrease in the third. Attainment of a given liquid throughput over time thus can be had by doubling ΔP and halving EFA (or vice versa). Although the influences of those two factors trend similarly, a given amount of change in one need not necessarily elicit a corresponding degree of alteration in the other. Furthermore, they are not equally convenient to handle. Once a filtration is in operation, it is far easier to vary the pressure than to exchange or augment the available filter area.
The interrelationship of filter area, ΔP, and time is constrained by the particle load. The time dedicated may be limited by considerations of product stability or by a need for restricting organism growth. As stipulated by US FDA regulation since 1960, prevention of grow-through is limited to an eight hour shift (4). A filtration may be slowed to increase the residence time of smaller particles within filter pores to promote their adsorptive sequestration, but increased residence time creates the potential for adsorption of formulation ingredients (e.g., the target protein itself). Durations can be set to meet scheduled staff or equipment availability.
Higher differential pressure levels expedite flow rates, thereby saving time. However, they may so compact filter cakes as to decrease their permeability—and hence the flow rate. In such cases, the outcomes depend on particle loads and their pattern of scattering over the filter (more precisely, on the ratio of solids deposition to the filter area).
Thicker filter cakes would seem more likely to be affected by compaction. The numbers, sizes, and shapes of suspensoids are seldom known—as also are their physical properties—so their resistance to deformation by differential pressures is also unknown. If they are gel-like or unstructured, elevated ΔPs may smear them into broader areas of filter fouling. Some users fear that particles made resistant to distortion by their rigidity could pierce filters when propelled by higher pressures. But we are not aware of such occurrences. That possibility is more likely when fluid has the low viscosity or low molecular density of a gas than with liquids.
By the same token, the reciprocity of temperature and viscosity can be manipulated to influence a liquid flow rate. Elevated flow rates, however, can create undesirable shear forces that could damage target proteins. High flow rates can also increase fouling of membrane or fiber structures within biological solution processes.
Generally, the purposes of a filtration can be met by relying on a more extensive effective filtration area. The more extensive that area and the more numerous its flow paths, the faster the possible rate of flow, the slower the rate of pore blockage, and the higher the throughput. Enlarged filter areas offer disadvantages as well. Countering their benefits are the cost of additional membrane area; higher extractables and leachables levels; and possible product loss by adsorption to the filter and/or in liquid imbibed by the filter and eventually discarded along with it.
Selection of Operating Conditions
In conjunction with the complexity of a filter’s pore structure for progressively decreasing and altering its porosity during operation, the quantity and physicochemical nature of suspended particles render a selected final filter susceptible to operational insufficiency. Available possibilities of optimal filtration operating conditions pose a conundrum. A general approach might begin with ΔP ≤30 psi (2 bar), and a sufficiency of EFA for a safety margin, coupled with the liquid’s low viscosity over a pertinent period of time. It is a compliment to the skill of pharmaceutical processors that they long ago empirically established a satisfactory and practical solution to this paradox: namely, relying on the technique of prefiltration.
Prefiltration depends on the use of larger filter areas. The selected final filter still performs its anticipated function, but the stress on its performance is reduced by another filter immediately upstream of it that assumes part of the particle load (Figure 1). The life of the final filter is prolonged by sparing it from an otherwise earlier blockage; thereby increasing throughput. The prefilter is sacrificed in assuming its portion of the load, but its expense is justified by the product gain it enables. Prefilters are usually depth-type in structure, thus costing less than the membrane filters they protect.
Prefilter use is simple; they need no integrity testing. Membrane filters with larger pore size ratings may, on occasion, serve as prefilters. But the more open porosity and nominal retentivity of depth filters allow permeation of a more generous portion of a fluid’s particulate mass. The final filter, as expected, retains the load reaching it. Its unaffected portion of porosity ensures an adequate performance. By assuming a portion of the particle load, the prefilter, in effect provides an EFA margin to the final filter that prevents premature filter blockage and obviates the need for midfiltration final-filter augmentation.
One of our purposes here and in Part 2 is to discuss, hypothetically, a filtration in terms of particles, pores, and liquid flows with a view toward understanding the structural requirements of prefilter and final filter arrangements that would fulfill the expectations of the size exclusion (or sieve retention) particle retention mechanism. Microporous membranes are characterized by pore size distributions. These filters are used in removal of organisms with particle size distributions. In most filtrations, neither the sizes nor numbers nor distributions thereof are generally known with regard to either the organisms or pores. And the foreseeable outcomes of filtrative sterilizations are additionally clouded by complicating influences of the filtration conditions. Nevertheless, what follows is our synthesis of several experimental findings reported by different investigators who challenged filters with dilute suspensions. We hope to reach an understanding of how pores respond to given particulate challenges.
A group of experimentalists in the Particle Technology Laboratory at the University of Minnesota made a very comprehensive investigation of the removal of latex spheres from liquid suspensions by membrane filters (5,6,7,8). Their impressive results have found practical application in semiconductor manufacturing, but they are likewise relevant to the needs of pharmaceutical processing. These several investigations used particle suspensions in dilute aqueous systems with surfactant molecules as ingredients. Particle concentrations (measured by particle counters) were ~20% those usually found in latex bead studies.
Table 1 details membrane retention of latex spheres. The middle column shows the total organism retention due to sieving and adsorption. The right hand column shows only sieving retention because the presence of surfactant eliminates adsorptive interactions. Thus, the total is reduced.
Table 1: Retention of 0.198-µm spheres by 0.2-µm rated membranes*
One study by Lee et al (7) led to conclusions analyzed by Zeman (9) that are a subject worth discussing here. The experimentation involved latex concentrations of 108–1010 per liter. Latex particles in each of five particle sizes (in aqueous suspensions containing 0.1% of a surfactant) were used to exert a continuous challenge against a 0.45-µm rated hydrophilized polyvinylidene fluoride (PVDF) membrane.
Composition of Latex Particles: The latex particles are made of polystyrene cross-linked by divinylbenzene. So they are hard, rigid, hydrophobic, noncharged entities presumably susceptible to adsorptive interactions—if at all only through hydrophobic adsorptions—and are incapable of hydrogen bonding with its attendant water wetting capability. Preparation by free-radical emulsion polymerization of styrene and a cross-linking agent involves dispersion by stirring of these hydrocarbonaceous ingredients into liquid spheres of increasing number and decreasing diameter in direct response to the vigor of stirring.
Those liquid droplets assume spherical form because their surface free energies, unengaged by water (the polar-medium surrounding them), are expressed as area-minimizing forces. Upon polymerization, the cross-linked liquid droplets thus transform into hard, incompressible solid spheres. This presents a “worst-case” scenario because spheres are the particle shape least easily retained by microporous membranes.
Cognizance of Adsorptive Sequestration: Although adsorptive interactions between particles and filter surfaces are understood, too many filtrations are investigated as if sieve retention were the sole means of particle removal. The next filtration investigated involved Escherichia coli thought to be arrested solely through sieving or size discrimination. Neglecting the contribution of adsorptive sequestration led to an incomplete analysis of experimental findings, and the predicted findings proved incorrect. Hypothesizing adsorptive retention rationalized the observed results. Perhaps this example of adsorptive bonding as an explanation may serve to highlight its pertinence.
The authors’ predictive model was based on a sieve retention mechanism, by which they calculated the log reduction values (LRVs) of different-diameter capillaries: 200 µm, 50 µm, and 5–20 µm. Observed and calculated retentions of test organisms by capillaries with 200-µm diameters were found to be in agreement with predictions based on sieving. Calculated LRV results for the 50-µm diameter capillaries underestimated retention of organisms on that very mechanism; actual retentions were better than those calculated. Calculated retentions were even less predictive for narrower pores (5–20 µm). The authors concluded that bacterial retention increased by some additional phenomenon “that we could not identify.” It would seem that added pore blockages yielded the greater retentions. The authors confirmed by scanning electron microscopy (SEM) that a clogging of pores
Adsorptive sequestrations would explain those findings. Particles (organisms) negotiating openings that are too large to be blocked by their passage may come close enough to the inner pore walls to become adsorptively fixed to them. A given size particle within a pore is more likely to encounter the wall of a more restricted pore diameter before convective flow transports it through. This would explain the better performance of the 50-µm rated pores and superior organism retentions of the 5–20 µm pores. Agreement with the prediction based solely on the size-discriminating mechanism confirms the relative absence of adsorptive influences in the 200-µm rated pore, which was presumably too large to permit much particle contact.
Surfactant Effects: Certain experimental trials by Lee et al (7) center on the sieve retention of latex particles essentially uncomplicated by adsorptive sequestrations. Addition of a nonionic surfactant to the suspension made that possible. The aqueous medium contained 0.1% Triton X-100 surfactant, a nonylethylene oxide adduct to a dodecylalkylated arylphenol from Rohm and Haas (www.rohmhaas.com). It is a sizable molecule with a molecular weight of ~375 Da and consists of a nonpolar aromatic substituent at one end with a chain of nine (on average) etheryl oxygens at the other.
The nonpolar ends of the surfactant molecules undergo hydrophobic adsorption with nonpolar surfaces of the hydrocarbonaceous latex particles, thereby enlarging them by coating the latex spheres. This increased in the distance separating each particle surface from the filter surface. The influence of this steric hindrance thus nullifies the adsorptive sequestration of particles. The result is enlargement of smaller particulates leading to their retention by sieving. This serves to alleviate the worst-case conditions, namely, small particle passage neither sieve retained nor adsorptively restrained.
It’s not surprising that Emory et al. found that different surfactant types—each with its own molecular features—vary in their influence on the retention of latex particles (10).
Part 2 will conclude this discussion by examining dilute particle suspension effects, particle size and shape, and filter pore structure, as well as further experimental findings.