Controlling temperature at small scale to mimic that of cold rooms or cold-boxes can be a challenge. The use of water baths
with column jacketing can provide precise temperature control, but the tubing leading up to the column can frequently act
as a mini heat exchanger, making it difficult to control the actual inlet temperature. Using heat exchangers on the inlet
tubing and column jacketing may be necessary for accurate, consistent, and precise temperature control.
The general approach towards qualification of a scale-down model is to run all operational parameters (inputs) at the center
of the operating range used at large scale (clinical or commercial manufacturing). The performance parameters (outputs) such
as step yields, pool purity, impurity clearance, elution volume and retention time should be similar across scales. Some of
the significant operational parameters include:
Figure 1. Chromatographic Profiles of Small- and Large-scale Columns. Raw material is a proprietary aqueous solution that
has undergone one step of preliminary purification. The small column (green data) is 1.6 cm dia by 12 cm. The large column
(black data) is 63 cm dia by 12 cm.
- Column loading (g protein/L resin) should be the same at both scales. In size-exclusion chromatography, the loading should
be based on a volume percentage of the column bed volume.
- Gradient lengths (in number of column volumes) and slopes (increase in percent limit buffer per unit time, or similar measurement)
should be the same for all steps of a chromatography procedure such as equilibration, wash, and elution.
- Temperature, pH, and conductivity should be identical for all buffers and feed streams. Differences in temperature across
scales can lead to significant changes in con- ductivity of some buffers that may impact step performance.
- Bed height of the chromatography column and the linear flow velocity are generally kept constant across scales. The ICH guideline
on viral safety states that the validity of the scaling down should be demonstrated, and that column bed height should be
shown to be representative of commercial scale manufacturing.1 If this is not possible and the bed height is different at small scale, the linear flow velocity should be adjusted appropriately
to keep the residence time constant.2
- Pooling criteria should be the same as in manufacturing. They can be online (conductivity or absorbance) or offline (gels
or HPLC). In our experience, if the pooling criteria are UV-based, it is a good practice to ensure that the UV detectors of
the small-scale systems correlate well with those in the large-scale system. UV detectors may be inaccurate or non-linear
over the optical density (O.D.) ranges required. If the absorbance range is non-linear at 280 nm, it may be prudent to measure
the on-line O.D. at 300 nm or some other wavelength that has a lower extinction coefficient.
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 using a t-test, tolerance intervals, or other statistical tests. Alternatively,
it may be demonstrated that the small-scale data are within the full-scale historical range. The approach taken may depend
on the inherent variability of the assays used to determine quality attributes or the amount of large-scale data available.
Sometimes there may be no large-scale data available at this stage of the project and the qualification of the small-scale
model may have to be done retrospectively after data from large scale are available. It is a good practice to perform several
small-scale runs (at least three), to better understand the inherent variability of the process, particularly when there are
limited amounts of large-scale data available for comparison.