Double Patenting over Published Prior Patent

Hoffmann-La Roche Limited v Sandoz Canada Inc 2021 FC 384 Manson J

2,667,654 / 2,709,997 / pirfenidone / ESBRIET / NOC

As I mentioned in my first post on this decision, Roche’s 997 patent was found to be invalid for obviousness-type double patenting over Roche’s 654 patent [148]–[153], even though the publication date of the 654 patent, June 26, 2008, was before the claim date of the 997 patent, November 10, 2008 [9], [11], [153], so the 654 patent was prior art over the 997 patent. In my first post, I didn’t make much of this point. I just said that I was “a bit puzzled” as to why the argument was run on a double patenting basis, given that the 654 patent was prior art over the 997 and so, under Hospira 2020 FCA 30, would have been part of the state of the art against the 997 patent. I suggested that “Perhaps Sandoz felt it was safer to use double patenting rather than to rely on the Hospira doctrine—presumably the 654 patent was not part of the common general knowledge and would not have been discoverable in a reasonably diligent search.”

On further reflection, it seems to me problematic to use double patenting in this situation. Now, it may not matter, if indeed under Hospira the 654 patent is part of the state of the art against the 997 patent. But what if there is a difference? Suppose the 654 patent could not be used in an obviousness attack against the 997 patent, but it could be used in a double patenting attack. In that case, allowing a double patenting attack would mean that the patentee is in a worse position than an unrelated third party. That is, suppose that the 654 patent can be raised against the 997 patent in a double patenting attack, but not in an obviousness attack, and suppose that the 997 patent is obvious over the 654 patent, but not otherwise. That means that the 997 patent would be considered a valid inventive improvement if granted to a third party, but would be invalid if granted to Roche. This seems to me to be contrary to the general principle that validity is objective. It also strikes me as unfair that Roche should be at a disadvantage in patenting improvements to a base technology that it invented and disclosed. 

There is also a problem of consistency with the Act. In Sanofi 2008 SCC 61 [12], the Supreme Court remarked that the law of patents is “wholly statutory.” That is a considerable overstatement: the double patenting doctrine itself, which was approved by the SCC in Whirlpool 2000 SCC 67, is a judicially created doctrine which, in the classic application, prevents an inventor from obtaining a second patent for an invention which is an obvious variant of an invention disclosed in its own unpublished prior application. The doctrine was thought to be necessary precisely because that prior application is not part of the state of the art as statutorily defined in s 28.3. There are many other instances where the courts have supplemented the sparse or ambiguous language of the statute: the non-obviousness requirement, fundamental as it is, was not codified until 1993. But there are limits. As the SCC also pointed out in Sanofi 2008 SCC 61 [12], “the general concepts which are the common currency of patent lawyers are founded on a statutory text, and cannot have any other firm foundation.” If the prior application has already been published, there is no loophole to close.

Now, as noted, none of this matters if, under Hospira the 654 patent could also be raised in an obviousness attack against the 997 patent even if granted to a third party. But that isn’t an adequate reason for holding that the double patenting doctrine applies even if the prior application is published. Hospira postdates the double patenting doctrine, and the two developed independently. If the double patenting doctrine applies now, it would also have applied prior to Hospira, when it did undoubtedly make a difference. Further, Hospira is recent and there is perhaps still some uncertainty as to its scope. Applying the double patenting doctrine in this scenario means that the 997 patent was held to be invalid without ever addressing whether the 654 patent was part of the state of the art against it.

Another problem with using double patenting in this case is that the obviousness-type double patenting was assessed as of the claim date of the 997 patent [153]. That is clearly the appropriate date for an obviousness attack, but it is not clear that it is the correct date for a double patenting attack, as it is currently unsettled whether the appropriate date is the claim date of the first (654) or second (997) patent: see here and here. Now, Manson J has previously held that the appropriate date is the second claim date, and perhaps that is why he used it, but the point is not settled at the FCA level, and it would have been preferable to have this issue discussed expressly.

Since Manson J did not have any discussion at all of whether it was appropriate to use double patenting in this situation, I would presume that Roche simply didn’t object to the use of the double patenting attack and instead chose to focus its efforts on the substantive question of obviousness. Consequently, I would suggest that this decision is not strong authority for the proposition that it is permissible to use obviousness-type double patenting when the first patent is already published.