Validity

This article is part of the latest edition of our Patents: Review of the Year publication.

The skilled person and their common general knowledge (CGK)

If a problem to be solved exists only briefly before a solution is found and patented, it can be difficult to establish the common general knowledge before the priority date. Was the problem established for long enough for knowledge surrounding it to be published? This issue arose before Meade J in Nokia v Oppo²⁶ and as the Judge explained, the identification of CGK was difficult in this situation because it was not possible to rely on the “classic way of proving CGK” by using well-known textbooks. Instead, the Court was faced with using multiple documents within which lay the danger that focusing on any particular individual document could be misleading. Overall, the Judge concluded that “What matters is information that became generally accepted; individual people in real life will have read different collations of documents from which they obtained the same information [which was CGK] … ideas specific to only one or two [documents] … were not”.

Novelty

When assessing novelty in Nokia v Oppo²⁷ the disclosure of a prior art citation known as Woo (a “very unclear” published European Patent Application), available for novelty only under s. 2(3)) of the Act, Meade J remarked on the limitations of expert evidence for novelty, noting that it is “admissible where it elucidate[s] the technical considerations relevant to understanding [a] document and inadmissible where it descend[s] into mere analysis of words. Attributing meaning (or lack of it) once the technical context has been explained is the Court’s function”. The Judge also dealt with the irregular situation whereby some parts of Woo claimed a priority date earlier than that of the patent in issue and other parts did not. Nokia argued that whilst only those parts of Woo with an earlier priority date could be relied upon for anticipation, the other parts of Woo (without priority) could be used to interpret the parts relied upon. Meade J thought that this was “intuitively rather odd” but in the end agreed with it since “nothing in s. 2(3) … says that only part of [a document] must be read for the purposes of interpreting it, and to not read all of it might give its individual parts a meaning which was not (objectively) intended”. However, the Judge also made clear that this nuance was not determinative of his conclusion that although it was arguable that Woo “covered” the method of the patent it did not “disclose” it and therefore did not anticipate it.

Obviousness

Continuing with Nokia v Oppo²⁸ in relation to obviousness, Oppo presented a stepwise analysis whereby the skilled person was assumed to seek to improve upon the cited prior art (ZTE) and find another document (LGE) which could be mosaicked with ZTE to render the patent obvious. Meade J noted upfront that, though a permitted approach²⁹, this presented acute dangers of hindsight following Technograph³⁰. Specifically with respect to the mosaic of ZTE and LGE, Meade J reminded himself that the skilled person does not “approach any particular citation with the expectation in advance that it will contain something useful”³¹. This was important because there was no express cross-reference between ZTE and LGE so Oppo’s case required the skilled person to read all of the relevant documents available to them (LGE was one of many) to improve the disclosure of ZTE. Meade J found that this amounted to an argument that all of the documents were CGK, which he rejected. He also rejected an argument that, since ZTE and LGE “were of a broadly similar kind it would be obvious to connect the one to the other”. Furthermore, since each document presented a “self-contained scheme”, Meade J concluded overall that “the skilled person would not feel on solid ground cutting and pasting them” together and, as such, Oppo’s obviousness case failed.

A similar stepwise approach to obviousness based on a sequential reading of prior art documents was the subject of appeal in Optis v Apple³², concerning Optis’ patent for a method by which mobile phones access information from the network. An important part of the method involved a technique for generating a pseudo-random number. Pseudo-Random Numbers Generators (RNGs) are not truly random, such that different RNGs may be better or worse (i.e. exhibit more or less randomisation). Apple’s argument at first instance was that the skilled person reading the prior art referred to as Ericsson would see a problem with the RNG it disclosed and seek to improve it. This would motivate them to look at a second prior art document (NRC3), which was CGK, and thence to a third document (Knuth), to which the second cross-referred, in which they would find the RNG used in the patented method.

The danger of hindsight in this analysis led to disagreement between Arnold LJ and Birss LJ over how Meade J had treated the expert evidence at first instance. NRC3 warned against using the type of RNGs described in Knuth but the Judge relied on evidence from Apple’s expert that the skilled person would not heed these warnings since the level of randomisation necessary for the problem considered by the patent was lower than that required in other technical fields. Arnold LJ did not consider this to be a conclusion open to the Judge since there was no sufficient reason in evidence for the skilled person to ignore the warnings in NRC3 and the skilled person would have to have reasons for ignoring the warnings based on their common general knowledge. Birss LJ dissented. Expert evidence on what a skilled person might think in light of a document is admissible and hindsight need not be the only explanation for taking a certain position. To explain this, Birss LJ used a culinary metaphor: a modern recipe book might look back with scorn on a 1970s recipe for making spaghetti carbonara with cream cheese, yet an expert would be entitled to explain that the modern book was aimed at the gourmet and those with less refined tastes operating in a hurry might actually value the cream cheese method. Therefore, to arrive at a recipe using cream cheese might not necessarily involve hindsight. Accordingly, he did not believe Meade J had erred when finding it obvious to proceed from NRC3 to Knuth. However, Nugee LJ agreed with Arnold LJ and the appeal was allowed with the effect that Optis’ patent was held valid.

The risk of invalidation by old or obscure prior art may seem harsh (see Floyd LJ in Hozelock³³), but examples of very old art killing a patent are relatively rare, not least because the age of a piece of prior art can be a factor that influences the skilled person. An almost 50-year old citation was almost successful in EnOcean v Far Eastern Manufacturing³⁴. However, whilst Nicholas Caddick KC (sitting as a Deputy) decided its age would not be off-putting (the skilled person must read any piece of prior art with interest and the question is whether the skilled person perceives in it something of relevance to the problem facing them) he also decided it would not have been obvious to change the workable solution presented in the prior art to something more complicated even though the concepts needed for the skilled person to arrive at the invention were part of the CGK.

The patent in Teva & Sandoz v Astellas³⁵, claiming mirabegron for the treatment of overactive bladder (OAB), had been held valid by Meade J at first instance in 2022, the obviousness attack having failed. The prior art disclosed the drug and some uses, but not the patented indication. In the context of a wider Markush formula, other members of the class were explained to be selective, and therefore useful in treatment, but uncertainty existed over the degree of selectivity and potency. 

On appeal³⁶, this “uncertainty” aspect gave rise to an interesting consideration of the extent to which the patent must demonstrate that it has arrived at a solution. In this case, the specification presented the results of experiments in rats but said nothing about success in humans. However, the appellants accepted it was plausible that the invention would work. That being the case, Arnold LJ pointed to Conor v Angiotech³⁷ as making it clear that the question of obviousness does not depend on the amount of evidence presented in the patent. He also dismissed an argument that the patent made no technical contribution, having neither identified a new human ß3-AR agonist (mirabegron having been disclosed as such in the prior art) nor identified a new use for ß3-AR agonists (their potential for the treatment of OAB being common general knowledge). Arnold LJ agreed with Meade J that the technical contribution lay in teaching the use of mirabegron in treating OAB, and giving specific, concrete results in identified assays, albeit not in humans or human tissue. Accordingly, there was no basis for interfering with Meade J’s judgment and the appeal was dismissed.

Another reminder that first instance decisions on obviousness are hard to appeal came with the upholding of Mr Campbell Forsyth’s decision in Advanced Bionics v Med-El³⁸ that Med-El’s patent was obvious. In giving judgment, Arnold LJ was firm in his view that Mr Forsyth had been entitled to find as he did and no errors of principle or law were committed. On the evidence before the Deputy Judge, the patent was bad, notwithstanding that it was later upheld by the Technical Board of Appeal (TBA) of the EPO. Different tribunals faced with different evidence are entitled to come to different decisions. This was also a case where the appellant patentee sought to rely on secondary evidence concerning long felt want: all the major companies in the field were said to have been aware of the prior art relied upon for many years yet not developed it. However, the evidence, a concession from the other side’s expert, fell short of proving that the citation was well known and in any event, even if it had been stronger, the evidence would have been secondary and not enough in this case to assist the patentee.

Plausibility

One of the year’s most eagerly anticipated judgments was from the Court of Appeal in the apixaban litigation, which dealt with plausibility in the weeks following the EPO Enlarged Board of Appeal (EBA) decision in G2/21. Arnold LJ was the first Judge in Europe to address his understanding of the decision in a judgment.

Readers will recall that in 2022, Meade J invalidated BMS’ compound patent for its blockbuster blood thinner apixaban (marketed as Eliquis®) for lack of plausibility³⁹. In upholding that decision on appeal⁴⁰, Arnold LJ rejected BMS’ contention that in the case of a claim to a single chemical compound there is no requirement that the specification makes it plausible that the compound is useful. In doing so, Arnold LJ was clear that the Supreme Court decision in Warner-Lambert⁴¹  was binding upon him and applies equally to compound claims as it does to second medical use claims. There is no lower standard for compound claims. He also rejected BMS’ proposition that a patent providing encouragement to test a particular compound in simple tests should be plausible, following the comments of Sumption SCJ in Warner-Lambert.

In G2/21, the EBA was asked to choose between two apparently conflicting lines of case law on whether to allow a patentee to rely on post-filed evidence to support a technical effect, referred to as ab initio plausibility and ab initio implausibility. The EBA considered the two lines of case law to be reconcilable under a unifying concept: “A patent applicant or proprietor may rely upon a technical effect for inventive step if the skilled person, having the common general knowledge in mind, and based on the application as originally filed, would derive said effect as being encompassed by the technical teaching and embodied by the same originally disclosed invention”. Arnold LJ disagreed that the two lines of case law were reconcilable; in Warner-Lambert the majority favoured ab initio plausibility and found the relevant claim invalid, whilst the minority favoured ab initio implausibility and found it valid. Furthermore, he considered the test set out by the EBA to be closer to ab initio plausibility than ab initio implausibility. However, given that he was bound by Warner-Lambert, this is commentary rather than law.

BMS appealed this decision to the Supreme Court, but permission was refused on the basis that the appeal raised no arguable point of law. At the time of this refusal, the Supreme Court was set to hear the appeal of the Court of Appeal’s decision in FibroGen v Akebia⁴² in March 2024, which, although concerned with slightly different issues to Sandoz, would have still presented the Supreme Court with an opportunity to comment on G2/21. However, FibroGen has now settled and so practitioners must await another opportunity for resolution of the apparent divergence between the English court (and the Warner-Lambert standard of ab initio plausibility) and the TBA, French and Dutch courts, which have since concluded that the EBA’s decision is far closer to the lower standard of ab initio implausibility. Patience may be required.

Insufficiency

Bookending the EBA’s decision in G2/21 and the Court of Appeal’s decision in Sandoz and Teva v BMS⁴³ (above), were two detailed considerations of plausibility in the context of insufficiency: first, Meade J’s decision in Gilead v Nucana⁴⁴ and then Mellor J's decision in Astellas v Teva and Sandoz⁴⁵. 

Nucana's two patents included product claims defined by a Markush formula. Both lack of industrial applicability and lack of plausibility were in issue, although ultimately the requirement for industrial applicability was relied upon to ensure that the technical contribution should not be set too low. Meade J rejected the “potential to be used in the treatment of cancer” as an acceptable target: “It seemed to me to amount to saying that the specification just has to render it plausible that the effect might or might not exist, which is meaningless”. Likewise, the patent did not provide enough information about the mechanism of action of the compounds for them to be useful as research tools. The Judge did, however, apply his observation in Sandoz and Teva v BMS⁴⁶ that a patentee does not have to rely on the most ambitious technical contribution the specification discloses, hence cytotoxicity (without demonstrable effect as a treatment of cancer) was a feasible contribution. Notwithstanding this finding, the Judge decided that the skilled person would be unable to predict which compounds within Nucana’s Markush formula would exhibit meaningful cytotoxic effects. As a result, and hedging his bets on the outcome of the EBA decision in G2/21 (published 2 days later), he held the patents to be invalid on both the ab initio plausibility test and on the ab initio implausibility test.

Lack of plausibility was not the only defect in Nucana’s patents – they were also manifestly insufficient on the classical basis (undue burden). In fact, Gilead v Nucana was closely similar in substance to Idenix v Gilead⁴⁷ ten years ago, in which Arnold J held Idenix’ patent invalid on this basis. The broad Markush formula of the claims in that case covered billions of compounds and Arnold J held that the patent was “setting the skilled team a substantial research project to select, synthesise and test the claimed compounds relying upon their own common general knowledge and claiming the results if they are successful”. Gilead’s pleadings advanced essentially the same case. Once again, the Court found there to be an undue burden for the skilled person to produce certain precursors for making some compounds covered by the claims. To hammer home the undue burden argument, Gilead relied extensively on secondary evidence comprising real work done in the field at the relevant time (so much so that the weight of material required a separate legal team to process). Meade J cautioned against excessive focus on secondary evidence but found that it confirmed his view based on the primary evidence that Nucana’s patent was invalid.

As mentioned above, the Astellas v Teva and Sandoz⁴⁸ case concerned a modified release formulation containing mirabegron, said to reduce food effect. Mellor J's decision provided an application of the framework established by the Court of Appeal in FibroGen⁴⁹ to deal with the question of whether it is possible to make a “reasonable prediction” that the invention will work with substantially everything falling within the scope of the claim. As readers will recall, Birss LJ did this by adding two antecedent questions or steps: (i) what falls within the scope of the claimed class?; and (ii) what does it mean to say the invention works? Once these have been determined, the reasonable prediction question (step (iii)) can be assessed more effectively. Mellor J applied Birss LJ’s categorisation of structural and functional features from FibroGen, finding that the dissolution profile claimed was a functional feature that was part of the definition of the claimed class (step (i)), while the functional feature of reducing food effect was relevant to determining what it means to say the invention works (step (ii)). He found it was plausible that compositions falling within the class would be capable of reducing the food effect. Even though the upper end of the limit for dissolution rate appeared to be arbitrary, that did not render the claims insufficient because products beneath that limit would work.

The FibroGen questions also featured in Meade J’s decision in Teva Pharmaceutical Industries Ltd v Grünenthal GmbH⁵⁰ in which Teva sought to clear the way of Grünenthal’s formulation patent relating to formulations of a solution for intramuscular injection of the steroid testosterone undecanoate. Grünenthal had sought to amend the claims of the patent in suit both conditionally and unconditionally but in each case the amended claim included ranges of components within the intramuscular injection solution. The description contained examples with one particular formulation. Noting that the debate before him was really about FibroGen step (iii) (reasonable prediction), Meade J concluded that in the circumstances of the case and on the evidence before him, the degree of predictability about the behaviour of the intramuscular formulation as claimed was “simply extremely low, verging on nil for significant changes to relevant parameters” and, as such, all the claims of the patent as proposed to be amended either conditionally or unconditionally were insufficient for lack of plausibility across their scope and therefore invalid. In reaching this conclusion, Meade J was clear that he was not ruling that patentees were not permitted to generalise from preferred embodiments but rather that the extent of permitted generalisation was dependent on the subject matter of the patent.

Excluded subject matter

Practitioners who find this area of law difficult will be unsurprised by Meade J’s comments in Nokia v Oppo⁵¹ that the “case law is very important in this area, and has a complicated history”. In relation to step 2 of Aerotel⁵² (“identify the actual contribution” of the claim), Oppo pointed to HTC⁵³ where Floyd J had said that the baseline against which the contribution of the patent would be assessed included “any item of prior art for a novelty attack”. Accordingly, in attacking the patent, Oppo sought to rely on Woo (prior art available for novelty only under s. 2(3) of the Act). However, Floyd J’s judgment also suggests (obiter) that novelty-only prior art should not be included because, being unpublished at the priority date, it was not part of the “real state of the art”. Meade J considered that “this is not a straightforward point” but in the end thought that Woo could be relied upon for the purposes of establishing the contribution of the patent. In making the point that novelty can lie purely in excluded matter, the Judge gave an example of “a piece of novelty-only art that disclosed all the claimed features of a patent claim except for a claim feature that the product be painted blue (an aesthetic choice clearly excluded by Art. 52(2)(b))”. Notwithstanding this finding (which Meade J recognised could be said to be in disagreement with Floyd J in HTC) the contribution of Nokia’s patent was not found to be excluded subject matter and the validity of the patent was thus upheld.

Sir Anthony Mann’s decision in Emotional Perception AI’s Application⁵⁴ is welcome news for those seeking to patent AI inventions in the UK and may mark a major shift in UK IPO practice. Sitting in retirement, Sir Anthony found that the UK IPO had erred in finding⁵⁵ a novel artificial neural network (ANN) implementing a recommendation system and characterised by its training method as being excluded from patentability as a “program for a computer… as such” under s. 1(2)(c) of the Act. The hearing officer had considered that the emulated ANN could not be decoupled from the software platform that supports it. The applicant appealed on the grounds that: (1) there is no computer program; and (2) if there is a computer program, the exclusion does not apply because the claim reveals a technical contribution.

In relation to the first point, the Judge accepted the patentee’s submissions that an ANN is not a program for a computer and should, in effect, be treated as a piece of hardware, irrespective of whether it was directly implemented as hardware or as an “emulated ANN”. Accordingly, he found that the subject-matter exclusion did not capture this feature at all. Secondly, and in any event, he found that the claimed system demonstrated a technical effect outside the ANN that would be substantial enough to avoid the subject-matter exclusion. He found that the selection underlying the recommendation was based on “technical criteria which the system has worked out for itself” and that the output thereby constituted a technical effect outside the computer, thus escaping the subject-matter exclusion (if it had applied). Following this decision, the UK IPO has suspended its guidance on the examination of AI inventions and stated that “patent examiners should not object to inventions involving ANNs under the “program for a computer” exclusion.” However, that is unlikely to be the last word: on 15 December 2023, the UK IPO confirmed that it had been given leave to appeal.

Added matter

If there is a risk that claims to a Markush formula might be overly broad, amending to narrower claims is not necessarily straightforward. Although the accepted wisdom is that the UK Courts are less likely than the EPO to find that amendments add matter, that did not prove to be the case in Gilead v Nucana⁵⁶. Nucana had put forward unconditional amendments to narrow the possible substituents at X and Y of the Markush formula from H, F, Cl, Br, I, OH and Me to only one option for Y and only four options for X. What was at issue, therefore, was deletion or selection from multiple lists, a problem with the amendments which the TBA had identified in its preliminary opinion in the ongoing opposition proceedings. The Judge reviewed the principles of added matter, emphasising the goal of preventing an unwarranted advantage to the patentee and preserving legal certainty for third parties. To accord the original filing date of an application to a selection-type invention which was not disclosed in the application (because in reality it was only made at a later date) would give an unfair advantage to the applicant. In Meade J's view, the EPO authorities supported the position that it was relevant to ask whether the amended claim presents a different invention with a new technical contribution. This did not replace the gold standard test but could be a “likely symptom of there being added matter”. Overall, on the facts, the new class defined by the amendments was significantly different from that originally claimed and Nucana’s position that every possible combination was disclosed was rejected.

Added matter by intermediate generalisation is a difficult ground on which to succeed. It was raised without success in Ensygnia v Shell⁵⁷, in Nicoventures v Philip Morris⁵⁸ and in Abbott v Dexcom⁵⁹. In the latter, Dexcom argued that the PCT application only disclosed the use of a “biased retention feature” that interacted with a “detent” in the applicator housing. Claim 1 of the patent included integers involving biased retention features that did not interact with a detent. Dexcom’s reasoning was, in part, that there was a clear functional or structural relationship between the snap and detent feature disclosed in the PCT application and the amendment to refer only to a biased retention feature added matter. The Judge held that the amendment, whilst a generalisation of the snap and detent feature, in fact taught the skilled person nothing new. The added matter attack therefore failed.

Continue reading the other sections in publication:

1. Claim construction and infringement
2. Validity
3. FRAND
4. Supplementary protection certificates (“SPCs”)
5. Procedural issues
6. Unitary Patent/Unified Patent Court

================================

Click here to view all references for Patents: Review of the Year