Sequential Unmasking of Expert Witnesses

Amgen Inc v Pfizer Canada ULC 2020 FC 522 Southcott J
            1,341,537 / filgrastim / NEUPOGEN / NIVESTYM

Wednesday’s post gave an overview of the facts and discussed the main substantive holding in this decision, namely that the invention was obvious. The decision also raises sequential instruction of expert witnesses as a way of improving the usefulness of “blinding” the experts.

Southcott J summarized the current law by accepting that “blinding is not a guarantee of reliability. Whether it is a sufficient reason to prefer the evidence of one expert over another depends on the details of the evidence in a particular case. [T]he effect of blinding is a question of relevance, reliability, and weight. It is not a doctrinal matter” [309] (citing 2016 FC 857 [59]).

Southcott J noted that some structural problems may arise that impede effective blinding. For example, effective blinding requires that the expert not have read the patent so as to avoid hindsight bias; but obviousness must be assessed from the perspective of a person skilled in the art to which the patent pertains, and it is necessary to read the patent to determine who the skilled person is.

In this case, Pfizer got around this problem in a manner that met with Southcott J’s approval. Pfizer had one expert read the patent and provide an opinion on the skilled person, and then the second expert was provided with that definition of the skilled person:

[318] This approach demonstrates a potential means of avoiding the concern about blinding creating a structural impediment to experts employing the required tools to formulate their opinions. If one expert provides an opinion defining the Skilled Person after reviewing the relevant patent, that definition can be provided to another expert, who can then give a blinded opinion from the perspective of the Skilled Person without having read the patent or the other experts’ opinions.

Southcott J remarked that while this methodology was not in itself a sufficient reason to prefer Pfizer’s experts over those of Amgen, “blinding is a factor favourable to the reliability of Pfizer’s expert opinions on the work that would flow from [the state of the art]” [320].

This calls to mind the concept of ‘sequential unmasking’ referred to by Arnold LJ in Akebia v Fibrogen [2020] EWHC 866 (Pat):

[36] This submission illustrates why it can be advantageous to try to instruct expert witnesses in sequence, first asking them about the common general knowledge, then showing them the prior art and asking them questions such as what steps would be obvious in the light of it and only then showing them the patent in suit. This is a procedure known as “sequential unmasking” in the psychological literature (see generally on this subject C.T. Robertson and A.S. Kesselheim (eds), Blinding as a Solution to Bias, Academic Press, 2016). The point of it is to try to avoid, or at least reduce, hindsight. In my opinion, it is desirable to try to minimise hindsight on the part of expert witnesses where possible. There is no rule or principle that experts must be instructed sequentially, however.

Arnold LJ went on to note that

[T]here are often real practical problems in doing so. To take just one obvious example, any discussion about the common general knowledge must start by identifying the skilled person or team. How is this to be done if the expert cannot be shown the patent? One way is to ask the expert to make an assumption, which they can check later when they see the patent; but that is not necessarily a perfect solution.

Another, arguably preferable solution to this problem, is that used by Pfizer here, namely providing the second expert with the first’s opinion as to the characteristics of the skilled person.

In any event, it seems that some form of sequential unmasking may be a real advance in terms of allowing experts to provide an effective yet properly blinded opinion.