The question of how many batches have to be run is a recuring question, not always easy to answer and there is a great temptation to just say three (3)!
Although industry has typically used three batches during the process performance qualification (PPQ) phase to demonstrate that a process is capable of consistently delivering quality product, this “rule of three” batches or runs is no longer appropriate and the regulations do not now prescribe the number of runs required for process validation activities.
The FDA just state that, “each manufacturer should judge whether it has gained sufficient understanding to provide a high degree of assurance in its manufacturing process”, The EMA guidelines state “a minimum of 3 production scale batches should be submitted unless otherwise justified”.
ICH Q7 (12.50) says “3 consecutive batches should be used as a guide, but . . . “
The past typical use of three batches has always seemed very much a “finger in the air” to me as how (statistically) confident can you be after only three batches – a statistician friend tells me that for 95% confidence and 90% reliability, 30 runs would be required, however that is neither practical nor cost effective.
So – how do you determine how many batches to run?
There are two main approaches:
a) Risk based.
This uses a comprehensive risk analysis to assess how much process variation risk remains after applying existing process knowledge and process design data.
This is performed by establishing or assessing the main sources of process risk, perhaps using such techniques as Failure Modes and Effects Analysis (FMEA).
Another method that can be used to identify and characterize the significant process factors and parameters is based upon the design of experiments (DOE). Again, this method can require a large number of pre-PPQ DOE experiments to demonstrate the significant and interactive factors and can require a relatively high number of PPQ batches to be run (two significant interactive factors would require 4 PPQ runs, while 3 interactive factors would require 8 PPQ runs etc.to cover the minimum and maximum values of each parameter in combination.
A risk method is not statistically based and may lead to a high number of PPQ batches required if more than 2 sources of variation are determined, and for that reason alone is probably not practicable for anything other than for simple processes.
b) Statistical / pre-knowledge based.
Where practical and meaningful, a statistical method of determining the number of batches is recommended, although there is no standard industry approach.
The statistical: based process relies on calculations targeting capability, tolerance intervals, or overall reliability of meeting CQA acceptance criteria.
The concept of this approach is that sufficient historical data from development, engineering or pre-PPQ runs is obtained to develop a deep enough understanding of the process so that you can have a high enough statistical confidence to be able to predict the behaviour of the PPQ batches. the PPQ runs themselves are then confirmatory runs of your predicted process performance. The number of PPQ runs can be as low as 2 runs if your % confidence level can be shown to be high enough.
The downside of this approach is that it requires a relatively high number of batches to have been manufactured and sufficiently well monitored to be able to build up this statistical “model” of the process prior to PPQ.
There is a third approach, – being a “structural” approach with the number of PPQ batches required being determined by reference to the process’s complexity, dosage form strengths, and number of equipment trains. This can include bracketing and matrix strategies and may involve separating groups of unit operations into separate PPQ protocols. However, bracketing and matrix strategies are difficult and time consuming to justify and not all regulatory bodies will accept such a method.
A statistical approach to determining the number of PPQ batches is often used in combination with risk-based or structural strategies.
So – to come back to the original question – how many PPQ batches are requires?
The number of batches should be based on the variability of the process, the complexity of the process / product, process knowledge gained during development, supportive data and the overall experience of the manufacturer. As such there can never be a single answer to this question but in general the number of three (3) consecutive would be the general answer and is the one usually chosen by the industry. Even this should be justified but I rarely see this done. These should be full size commercial batches, however some process validation studies may be conducted on pilot scale batches if the process has not yet been scaled up – but again a justification of this approach is required.
However if you have extensive process development history and data and can write a well thought out justification then less than 3 batches may be acceptable.
If you have little or no prior knowledge of the equipment, process or product then it would be wise to plan on more than 3. Likewise, if you have a complex process or are looking to use a matrix style approach then the likelihood is that you will need significantly more than 3.