As with the chromatography steps, use feedstock from large-scale GMP manufacturing runs (or equivalent). Buffers should either
be made in GMP manufacturing or according to the manufacturing recipes.
Table 2. Comparison of Small-scale Disc Format with Large-scale Cartridge Format.6 Aqueous solution feed for a NFF step for
The retentate path-length of the equipment should be the same at both scales. The Millipore Pellicon XL format scales from
50 cm2 to 2.5 m2 and the Pall Centramate scales from 0.1 to 50 ft2. The type of pump used should be of the same shear characteristics. If a low-shear rotary-lobe pump is used at large-scale,
do not use a high-shear gear pump at small scale.
Select membranes from a lot that is representative of full-scale production. For new membranes, it is a good practice to utilize
normalized water permeability (NWP) as a benchmark for their integrity. Before running the system, run a baseline of the permeate
at 280 nm and a pressure-hold test to ensure the system is not leaking.
When making tests, run all operational parameters at the center point of the operating range used for the large-scale filtration.
These include: retentate cross-flow rate, transmembrane pressure (TMP), both inlet and outlet pressures, temperature, number
of diafiltration volumes, and concentration factor. Crossflow and transmembrane pressure are the two main drivers for removal
of low molecular weight solutes and concentration of the protein. Also make sure the membrane loading (g product/m2 membrane) is the same across scales. Adequate stirring of the retentate reservoir is important to ensure the diafiltration
is as complete as possible.
Pre-determine the acceptance criteria prior to the scale-down runs. These can be based on a statistical comparison of the
small-scale data against the full-scale data. Be cautious — as the system dead volume is almost always larger at the small
scale and will result in lower pool concentrations or lower step yields. Deviations such as these should be kept under consideration
when evaluating results from the scale-down model and also when setting the acceptance criteria.
TMP vs. flux curves are useful to compare the decay in flux over processing time. In cases where the UF or DF step affects
product quality, a measurement of these key quality attributes should be included.
Diafiltration efficiency is important for DF applications and should include measurements of diafiltration volumes vs. conductivity
and pH. A measurement of excipient concentrations may be appropriate for final formulation steps.
Figure 3 illustrates a scale-down of a depth filtration step. The 23-cm2 model closely mimics the performance at 1.8 m2 in the operating range with respect to the pressure drop and the filtrate turbidity vs. throughput.5 Table 2 shows data supporting the use of filter discs for scale-down modeling of large filter cartridges using the ØX174
bacteriophage model.6 Flux, yield, and viral clearance are shown to be comparable across the scales.
Creation of scale-down models that can meet the qualification requirements can be challenging. In general, downstream unit
operations are easier to scale down than upstream ones. Exceptions to this are protein-refolding unit operations that may
require more extensive modeling efforts.7 These models are useful for performing experimental studies in an efficient and economical fashion.