In my first blog post on the 4th PDA Europe Workshop on Monoclonal Antibodies, I touched upon some of the issues discussed around the QbD paradigm, the assessment of critical quality attributes and what this actually means to industry and regulators alike. There was plenty of discussion and debate, with one or two people questioning whether a QbD approach was even practicable.What was clear though is that not everything is clear! The muddy nature of things leaks through into my second blog post on sequence variants…

The session on sequence variants was introduced by Kathleen Francissen of Genentech, who asked, “What is acceptable and what is not”; as analytical methods approach the inherent “error rate” of biosynthetic processes and the actual risk of sequence variants is somewhat unknown, the answer to that question is a moving target.

The first presentation on sequence variants by John Stults, also from Genentech, considered both angles: detection and subsequent risk assessment. Detection of sequence variants, which are unintended amino acid substitutions caused by mistranslation or mutation, is not new. But traditionally, detection has occurred at the cell line development stage during sequence variant screening (if a sequence variant is detected, the cell line is abandoned). Importantly, detection capability is increasing; because of the discovery of a sequence variant late in Phase III development, Genentech implemented earlier systematic sequence variant analysis on both normal and extended cell age samples. If a sequence variant is discovered during extended characterisation, risk assessment is performed by a cross-functional team based on several principals, not least risk of immunogenicity.

There are several issues with measuring immunogenicity risk though. One is the limited scope of in silico assessments or animal models; another is the inability to measure low frequency immune responses in the small patient groups of Phase I/II studies.

So what is the regulatory perspective? Chris Holloway of ERA Consulting Group provided some extreme examples. In one instance, a client was considering the impact of a sequence variant discovered at Phase II. They were told it was certainly a concern – especially as the variant occurred at a level of 45%! He went on to point out though that generally (and presumably at much lower levels!) there was very little experience relating to the actual impact of sequence variants. Whether the risks are real or perceived, impurities are a fundamental quality attribute, and Holloway pointed out the need for batch-to-batch consistency.

In concluding, Holloway stressed the need to start the analytical methodology early on to detect sequence variants, memorably stating, “if only stock prices were linked to analytical method development, we would be a lot better off!” Much laughter ensued. Along with long list of recommendations to reduce regulatory risk, he stated that perhaps the single most important point was the need to fully investigate all factors that could conceivably affect sequence variant levels during process development.

A last key recommendation, backed by a case study, was to secure retained samples for future discoveries; it could alleviate regulatory concerns if a new variant is retroactively discovered in a legacy product.

When asked by a delegate how many batches were required to prove consistency, with reference to the oft-discussed three-batch minimum. Holloway pointed out the fallible nature of human pattern prediction by asking us to consider data from three batches: 20, 40, 30 ppm – ok, looks good. Now consider 20, 30, 40 ppm – is the process drifting? And 40, 30, 20 ppm? The team is improving…! Perhaps three-batch validation really is obsolete…

In the panel discussion that followed, there was more debate yet few solid answers over the risks of immunogenicity. But what everyone appeared to agree on was that if any level of sequence variant is found, extensive investigation is required; regulators will demand that sufficient data is presented to alleviate risks – whether real or perceived.