Really always three batches?

From the roundtable discussion on ”Process Validation in Pharma” on March 11^th, 2021

All good things come in threes

Does this also apply to validation batches? Starting point of our roundtable: The traditional three batches for validation is often too many and sometimes too few. The discussion around the traditional three batches is the tip of the iceberg.

Do we determine now what about the process is actually quality-critical for the product? And what risks arise from process variation of such critical parameters? What additional risks lurk in the procurement and sourcing processes? And how do we determine the required number of validation batches based on this analysis?

Is it just a scientific tradition behind the three? In any case, leaving this behind is symbolic of a new work culture around process validation.

It does not make sense

The Rule of thumb of three makes no difference how demanding the validation is: Is it a completely new process for a completely new product (e.g. production for a new vaccine), or is it just a product transfer, or the 200th validation for an in-house sterile filling process? It cannot make sense to apply the same effort to validation batches in all cases!

Process knowledge and systematic validation bring product safety, compliance, launch speed, and especially fewer deviations. So this can pay off very quickly.

What are the authorities demanding?

According to the FDA and EMA, in traditional validation, 3 batches may be sufficient for comprehensive validation. The required number of batches should be derived from the risk analysis. In particular, the FDA requires that not only is the process to be validated, but also the process performance. Today, there is an actual statistical rationale for only a few process validations. Many companies still work with the golden 3 batches. But approaches to statistical, risk-based determination of sample size for validation batches are evolving. This is especially the case when the cost of active ingredients is high or the time for development is short.

Good Practices: The early bird catches the worm

The earlier in the process development validation is considered, the better. This was the starting point for various discussion points:

1. Collect data before starting with GMP
   Interesting question: “Which is actually batch no. 1 of the 3?” The build-up of process knowledge naturally starts with starting the development process. This takes place long before production under GMP conditions. And this can be utilized and even further expanded. Existing data from the development can be used to statistically determine the validation batch quantity.
   The earlier problems are detected and corrected, the more easily and more profitably they can be eliminated. This is especially true before production under GMP conditions. Otherwise, problem solving almost always leads to investments in equipment, tools, extra tests…
2. With DOE in the direction of QbD (Design of Experiments/Quality by Design)
   Design of Experiments (DOE) is an especially good method here: As long as production is not yet under GMP conditions, one can learn a lot about the parameters and go into validation with good basic knowledge. The statistical basis for determining the validation batch quantity is one of the results, the correct setting of the process parameters critical to success is the other.
3. Defining the CQAs (Critical to Quality Attributes)
   Successful definition of the CQAs (Critical to Quality Attributes) requires in-depth expertise. This is exactly where the best experts from all areas are required.
   Every mistake made at this point costs a lot of time and money later on. Example? This is quite short.
4. The foundation: Identifying the risks
   The processes to be validated can be very different and this also applies to the underlying risks. Even today, processes are validated that pose no risk to the product at all, or validation takes place without identifying the risks. The reason for this is often that the process development department does not work closely enough with the process validation department.

Especially for the identification of risks, an overarching cooperation between R&D, dossier development, QA, QC, supply chain, and production is essential. Only in this way can those likely unconsidered risks arising from process steps also be identified.

Tools help! Not just statistics, tools…

Especially in process validation, one typically thinks of the dreaded statistical tools immediately. However, often enough, even the statistician cannot unleash his power: He has to consider so much variability and uncertain knowledge that the derivable efforts would go through the roof. This is where other tools help to sharpen the focus. Further, authorities are demanding tools for more transparency, for simple, low-effort interdepartmental cooperation, and for the development of robust processes. Good results have already been achieved in many companies. Good results have been achieved with Makigami to set-up the new processes.