For many years the "magical three validation runs" completed successfully were regarded as state-of-the-art in order to be able to define a process as validated. But already in the PIC/S document PI 006 on the topic qualification/validation (originally dated from 1996) is stated analogously that theoretically the number of validation runs should be sufficient to show the normal level of variations and to recognise trends and to collect sufficient data for the assessment. But nevertheless the number three is still mentioned in this document. In the new Guidance on Process Validation from the year 2011 the FDA continues with these reflections and does not mention a number of validation runs any more. Instead process understanding and knowledge are required.
Now the question arises for the pharmaceutical industry how many validation runs must be carried out today in order to show process knowledge and understanding.
Statistical approaches (such as by means of Bayesian statistics) on the basis of data of development seem to be very complicated and lead in some cases to very high numbers of validation runs. Data mining could be an alternative approach to show process knowledge and understanding. Then the defined number of data values also makes part of the calculation of process capability indices. A lecture during the 5th European GMP Conference taking place on 6 and 7 June 2013 in Heidelberg will illustrate what such a concept could look like.
PS. As attendee at the GMP-Conference you will receive the ECA Good Practice Guide Process Validation and the ECA GMP Matrix as free add-on.