Designing quality into a purification process also requires considering risks that may arise from the processing materials
themselves. Risks include protein ligands, leachables or extractables, and processing additives such as detergents. Also,
the process should be designed using only materials suitable for manufacturing according to current good manufacturing practices.
One of the most important factors for designing in quality is the ability to clean and sanitize packed columns and ancillary
Characterization and Robustness Studies
Once a suitable process has been identified, the next step toward achieving a validated state is to perform studies to quantify
cause–effect relationships from the inputs to the outputs of the process. DoE is a powerful statistical tool for quantifying
these relationships, but it is important to point out that any DoE study should be built on a foundation of process know-how
and empirical knowledge whenever available.
The final goal of many development studies of chromatographic unit operations is to establish ranges for critical process
parameters within which the process outputs meet acceptance limits. Generally, DoE studies leading up to this can be divided
into three categories, which are often performed in a sequential manner. The categories are:
Screening studies, in which a large number of process inputs are studied in a systematic way to identify the inputs that have the most significant
effects on the process outputs.
Optimization studies, in which the most important process inputs from a screening study are evaluated in more detail in order to quantify the
cause–effect relationships between process inputs and outputs.
Robustness studies, in which often a fairly large number of process inputs are studied in a systematic way, but with much smaller variation
intervals compared to those used in screening and optimization studies. Typically, the process inputs are varied in a systematic
way within their control limits to verify that the resulting process outputs are robust.
DoE studies can be performed at any scale, but due to time and cost restraints, screening studies are commonly performed at
laboratory scale, whereas optimization and robustness studies are performed at laboratory or pilot scale, or in some rare
cases at production scale. This varies, of course, between different processes and applications.
Data from a DoE study7 on the Capto S cation exchanger (GE Healthcare, Chalfont St. Giles, UK) will be used to illustrate the use of DoE from a
validation perspective. The effect from the process inputs residence time (2–6 minutes), conductivity (5–15 mS/cm), and pH
(4.5-5.5) on the process output dynamic binding capacity (QB 10%) for a monoclonal antibody (MAb) was studied. A total of
17 experiments were performed to quantify the effect of three process inputs on the process output.
The rather complex model coefficients for the effects of conductivity and pH on the dynamic binding capacity translate into
an easily interpretable response surface, as shown in Figure 2.
As shown by Figure 1, it was found that within the investigated ranges, residence time had a small effect compared to conductivity
and pH, whereas both conductivity and pH were shown to have significant linear as well as second degree curvature effects
on the QB 10% for the MAb. In addition, a significant interaction effect between pH and conductivity was found.
Figure 2 shows the combined effect from variations in conductivity and pH on the QB 10% for the studied MAb at a 95% confidence
level. Assuming that a dynamic binding capacity of at least 120 mg/mL is always desired from this process step, it would be
reasonable to set the target for pH at 5.1 and the target for conductivity at 6 mS/cm (as illustrated by the red dot) in order
to give some room for variation (illustrated by the blue lines) in these parameters and still be able to have a dynamic binding
capacity of at least 120 mg/mL.