by Giuliana Miglierini
There are many steps within a pharmaceutical production that may require the availability of a model of the manufacturing process in order to run targeted simulations. To this instance, a useful approach is represented by the so-called “small-scale models” (SSMs, or “scaled-down models”), that are usually developed to reflect the real working parameters available for a certain large manufacturing facility.
A small-scale model needs to undergo a process qualification (SSMQ) in order to be acceptable from the regulatory point of view. The main features and criticalities of SSMQ have been discussed in a series of articles published on BioProcess Online, and based on the results of a survey run between the representatives of large biopharmaceutical companies participating to the BioPhorum Development Group. A white paper on SSMs is also available.
The main requirements for an SSM
A critical requirement for a small-scale model to be accepted by regulators is its ability to exactly replicate the large-scale manufacturing process. This can be assessed and justified by choosing appropriate process parameters to be used as inputs for the simulation and obtaining outputs showing performance and quality attributes comparable to the large-scale process.
Small-scale models can be used both in early development, for example to support clinical manufacturing, and in late-stage development (e.g. to identify critical process parameters).
The overall quality of the model increases in the passage from early- to late-stage applications, due to the increasing number of data available to simulate the processes. Alternatively, a scientific evaluation of the process without application of a formal statistical method might be used, but a good experience and sufficient platform knowledge is needed in order to obtain valid results.
Other examples of the utility of SSMs in biopharmaceutical manufacturing include media stability and cell line stability studies, qualification of raw materials, impurity clearance validation, postapproval process changes and resolution of deviations.
The clearing of infectious viruses is a particularly critical step in biomanufacturing, and it should be run according to the ICH Q5A8 guideline; to this instance, SSMs may turn useful to validate the process at the laboratory scale. Other points to be kept in mind refer to the possibility of different layouts, mode of operation, geometry or materials for the systems used in small-scale vs large-scale plants.
Validation and qualification of the SSMs
A risk-based assessment of the parameters of choice can be used to validate the representativeness of model, with key performance indicators (e.g., titer, VCD, etc.) and product quality attributes (PQAs) used to run the comparison. A risk-based approach should be the choice also for the design of the small-scale model, taking into consideration both technical and business risks.
More than just one large-stage run (with a minimum of 3) is suggested to support the full qualification of the small-scale models by statistical analysis, according the survey. The choice to assess or qualify the SSM depends on its intended use.
The dimensions of the model can vary according to its specific target use. A benchtop-scale (1 L to 10 L) is common for upstream unit operations, but micro-scale bioreactors (15 to 250 mL) and pilot-scale (50 to 200L) models are other useful options. The benchtop scale of a chromatography
column can be used to model downstream processes, with micro-scale models or pilot plants as other alternatives. The article also reports a table to help identify the correct choice of the scale-independent “scaling parameter”.
In some instances, it might be advisable to use the same media and buffers as in the real manufacturing process, as well as the same raw materials. Procedures to prepare the buffers and other materials should be also comparable.
The BioPhorum Development Group provided examples of how to address qualification, including a satellite or non-satellite approach for upstream unit operations according to the characteristics of the inoculum transfer and scale of the run, the location of the development laboratories and the commercial site. An important parameter to be considered is the temperature for shipping, should it be required a transfer of materials between different locations; shipping at ≤-65°C is the preferred choice for many companies, writes the authors.
Different procedures for filtration have been also addressed, as well as the analytical setup for small-scale experiments; measures may be run in the QC GMP laboratories associated to the manufacturing site or in non-GMP labs for small-scale model qualification. A mix of the two may represent the preferred option in many cases, indicates the article. Training is fundamental to ensure the consistency of small-scale unit operations independent of the operator. Formal documentation should be also produced should the small-scale model undergo new runs of qualification.
The choice of the statistical methods
All data obtained both from the small-scale model and the large manufacturing plant needs to undergo a statistical analysis to be used for the qualification of the production process.
Descriptive statistical methods may depend upon the satellite or non-satellite character of the study, and they may turn useful to provide data in the form of scattered plots to be used for qualification assessment, for example by SMEs or health authorities.
Inferential statistical methods compare data obtained from the small-scale model and the atscale one, which must be representative of populations and referred to stable processes all over the product lifetime. Attention should be paid to the indication of “equivalent” or “notequivalent” results obtained from the applied method, as errors are possible in the 5-10% of cases.
“This is an important fact often overlooked by scientists and health authorities in evaluating the statistical component in a qualification report. It is also an important rationale for not using statistical methods alone to qualify or not qualify a model”, warn the authors of the article. Possible examples of inferential statistical procedures are the difference tests (or null hypothesis significance tests, NHSTs) known as T-test and F-test. Equivalence tests (Two One Sided T-tests, TOST) are also possible to obtain evidence of equivalency, especially in the case of a satellite design of the experiment. Quality range (QR) methods are another available option, useful to establish the population ranges. Multivariate analysis (MVA) provides the possibility to consider different, time-based data sets simultaneously, thus supporting the study of the processes under a time evolution perspective.