Trapped in the examples: Escaping intermediate generalisations in pharma patents

Intermediate generalisations are a familiar source of added matter objections at the EPO. An intermediate generalisation is what you get when a feature is lifted out of a worked example and put into a claim without the rest of the example. The claim ends up neither as broad as the general disclosure nor as narrow as the example, but somewhere in between. This is particular problem in pharma and biotech where our examples often describe complete products, whether a CAR construct delivered by a particular lentiviral vector, an LNP with a fixed combination and ratio of lipids, a formulation or a dosage regimen. Every exemplified embodiment will also almost always contain features that the inventors know are not essential or may be changed. Demonstrating that each of these features is not an essential component of the invention would mean running a control for every vector element, lipid, excipient, dose and process step (generally way beyond most R&D budgets and timescales). However, without this data, explaining to a Board of Appeal why a feature used in every example can safely be left out of the claim can be an uphill battle.

Looking at recent Board of Appeal decisions on this issue in pharma and biotech, many cases did not end well for the patentee. The recent case law shows just how unforgiving the gold standard can be. Nonetheless, sometimes a lucky patentee escapes and there are some important lessons that may be extracted from these cases.

Legal background: Intermediate generalisations

Added matter at the EPO is governed by Article 123(2) EPC. An amendment is allowable only if the skilled person would derive the amended subject-matter directly and unambiguously, using common general knowledge, from the whole of the application as filed (G2/10). This is the gold standard. The EPO Guidelines (H-V, 3.2.1) add criteria specific to intermediate generalisations. A feature may be extracted from a specific embodiment only if it is not related to, or inextricably linked with, the other features of that embodiment, and only if the overall disclosure justifies its isolation. These criteria are, however, only an aid. They are necessary but not sufficient (T 321/23), and meeting them does not replace the gold standard, which remains the test.

For an intermediate generalisation, the difficulty is that an intermediate combination of features may never actually have been described in the application as filed. If the skilled person could not derive it directly and unambiguously from the application, the amendment is considered by the EPO to add matter.

This is very much a European problem rather than a US one. The USPTO has a written description requirement and a bar on new matter, but it generally applies them far more leniently to this kind of amendment. Applications drafted in the US with US practice in mind are therefore a large part of the problem. They tend to describe embodiments in exhaustive detail without ever saying which features matter and which are optional. Given their reliance on data, biotech and pharma patents are particularly exposed to this problem.

Bratislava

The biotech and pharma generalisation problem

Intermediate generalisations are a particular risk in biotech and pharma. The invention in these fields is rarely a single, self-contained feature. Instead, it is typically a sequence, an antibody defined by its structure and its function, a multi-component cell therapy or delivery vehicle, a formulation, a dosage regimen or a combination of drugs. Each is usually made, tested and described as a whole.

Biotech and pharma inventions also depend heavily on data. Functional claims in particular must be backed by experiments. Examiners and Opponents therefore look to the examples first when deciding what the invention is and how broadly it can be claimed. However, the more specific and complete the examples, the more features there are that the applicant may later wish to leave behind. A feature that appears in every example, but that the description never discusses on its own, can look inextricably linked to everything around it, even where the inventors know that it is not essential.

The risk is greatest where the examples describe a complete product or process that the applicant may later want to claim in parts.

Intermediate generalisations and cell therapy

The first, and perhaps the most obvious, is the increasingly complex biological product. A CAR-T construct is made up of an antigen-binding domain, a hinge, a transmembrane domain, one or more costimulatory domains and a signalling domain, delivered by a particular vector. An LNP combines an ionisable lipid, a helper lipid, cholesterol and a PEG-lipid in particular molar ratios, with a particular payload. The examples will typically test one or two complete constructs or formulations. If the applicant later wants to claim the new binder or the new ionisable lipid on its own, it must find a disclosure that presents that component independently of the rest of the construct. Gene therapy and mRNA products raise the same issue. An AAV transgene is usually tested with a single capsid, promoter and set of regulatory elements, and an mRNA with a single cap, set of UTRs, poly(A) tail and nucleoside modification.

Intermediate generalisations and antibody platforms 

The same problem arises for antibodies. CDRs are frequently disclosed only in the context of a particular framework, isotype or set of Fc mutations, and tested in a single format. Claiming the same binding domain in a bispecific, an ADC or a CAR may then require a basis that the application does not provide. Functional features also carry a risk. An affinity, an IC50 or a level of cytotoxicity is only meaningful under the assay conditions in which it was measured, and lifting the value into the claim without those conditions risks a further intermediate generalisation.

Intermediate generalisations and clinical inventions

Doses and clinical features are another hotspot of intermediate generalisation issues. The data in the application are typically for specific doses in specific indications, often on specific schedules (IPKat). If the dose eventually selected for the clinic is slightly different, or the applicant wishes to claim the dose separately from the indication or regimen with which it was disclosed, it may struggle to find a basis. Features taken from a clinical trial protocol raise a related problem. A claim drafted around the trial's patient population, inclusion criteria, premedication, or dosing schedule may no longer cover the product once subsequent data lead to a different approved label. Removing those features after filing may then be objected to as an intermediate generalisation. The same applies to combination therapies tested only in a particular sequence, in a particular line of therapy, or in a patient population defined by a biomarker cut-off measured with a particular assay.

Intermediate generalisations and formulations

Formulations and manufacturing processes raise similar issues. A formulation may be tested only with a specific excipient or buffer that is not actually essential, and that a competitor could easily substitute. Convincing an Examiner or a Board of Appeal that the excipient can be omitted is difficult where every example includes it. Similarly, cell therapy manufacturing and differentiation processes are typically described as a single protocol, in which one step, time point, growth factor or culture medium is always used. A competitor may easily design around that feature, but the applicant may have no basis to claim the process more broadly (IPKat). 

Is it actually a generalisation?

It is first important to define what is and what is not an intermediate generalisation. T 630/24 concerned a University of Tokyo application for compositions of gut bacteria that induce regulatory T cells (IPKat). The application disclosed bacteria belonging both to the genus Clostridium and to Clostridium clusters IV or XIVa. The Applicant deleted the genus and kept the clusters, and argued that this was a justified intermediate generalisation because the taxonomic and phylogenetic classifications had no clearly recognisable functional or structural relationship and were not inextricably linked.

The Board of Appeal did not even engage with this argument because it considered that the amendment had been misdescribed and was not in fact a generalisation. Clusters are a broader phylogenetic group that spans several taxonomic genera. Therefore, dropping the genus was considered to widen the claim rather than leaving it somewhere in the middle as an intermediate generalisation. The question of whether the two classifications were functionally or structurally related therefore was considered not to arise.

This case shows us that an argument built on the intermediate generalisation criteria presupposes that the amendment really does sit between a general disclosure and a specific one. Where it does not, these arguments are then addressed to the wrong question and can put you on the back foot with respect to your patentability defence. This also shows us how complex added matter objections can be in biotech.

Extrapolating function from structure

Where an example is the only basis for a claim, the focus of an added matter assessment is on the purpose of the example and how much can be extrapolated from it. In Europe, it is possible to claim broad platforms of biotechnology based on function, and there is therefore a temptation to extract function from the examples in order to avoid narrow structural limitations. However, if you didn’t plan functional limitations in your original application, intermediate generalisation is a risk (IPKat).

T 660/24 concerned Chugai's EP 3130606, directed to cytotoxicity-inducing bispecific antibodies. Claim 1 required two Fab domains with defined specificities and an Fc region with decreased Fcγ receptor binding. The examples disclosed highly specific IgG-based constructs relying on heterologous heavy chain association and isotype-specific constant regions. The Board of Appeal found that the claim had taken the functional idea from the examples and discarded the necessary structural features:

"Claim 1 generalises the antibody constructs disclosed in the examples by abstracting only the functional concepts ... from the structurally and functionally highly specific IgG-based synthetic constructs used in the examples. In these constructs, the structural features and the associated functional properties are interdependent in the context of the bispecific antibody as a whole. This constitutes an unallowable intermediate generalisation" (r. 9). The patent was revoked.

A similar fate met the CAR-T vector claim in T 250/24 (IPKat). Citing T 895/18 and T 1362/15, the Board of Appeal in this case noted that neither claim structure nor the presence of broad embodiments suffices to support an intermediate generalisation "unless the specific combination of features can be directly and unambiguously inferred from the disclosure of the application as filed" (r. 4.1.2). Three selections from lists with no pointer were needed to arrive at the claim, namely the BCMA target, the CD8α hinge and the CD137 costimulatory domain (r. 3, r. 7). In addition, the omission of features present in the example constructs produced "an undisclosed intermediate generalisation between a general and a specific disclosure in the earlier application as filed" (r. 7).

Along the same lines, in T 1535/23, Pfizer's divisional EP 3431475 claimed a method of making crystalline palbociclib free base characterised by its X-ray diffraction peaks and a D90 particle size but omitted the specific surface area that the parent claim had required. Every passage relied on by the Patentee for basis described embodiments having the omitted surface area. Example 7 also characterised the form by all three parameters together. The Board of Appeal thus found that nothing in the application as filed taught that the surface area could be dispensed with and the patent was revoked.

Numbers are only as good as what they are attached to

Selecting and combining specific numbers is also tricky from an added matter perspective. In T 1365/24, the applicant, Novartis, claimed eltrombopag olamine tablets with an X90 particle size of 41 to 68 micrometres. The description of the parent disclosed only ranges of 10 to 90 and 20 to 50 micrometres, mostly in passages tied to a particular process and set of excipients, or to capsules. The claimed values existed nowhere except as points read off the dissolution curves in a Figure. The Board of Appeal thus found that the values had been "isolated from a specific context and selected among several possibilities from the data of Figure 2" (r. 1.4.1), and that the dissolution data were "the result of several features inextricably linked together" (r. 1.4.2).

T 1462/24 concerned Biogen's EP 2653873 and the 480 mg daily dose of dimethyl fumarate for multiple sclerosis. The dosage paragraph of the earlier application gave a series of ranges and named no disease. The claimed 480 mg was only the lower end of a less preferred range, and the paragraph in fact pointed to 720 mg (r. 9.10). The Board of Appeal found no pointer linking those dosages to multiple sclerosis, and rejected the Patentee's attack on the selections from lists methodology, holding it consistent with G2/10: "Indeed, direct and unambiguous disclosure cannot be equated to subject-matter resulting merely from a combination of possibilities selected from a reservoir of equally applicable options or alternatives that define a broad technical field" (r. 24). The patent was revoked, and costs apportioned against the Patentee.

Relying on a general vague passage

The case in T 1103/23 is also worth a mention. This case related to duplex sequencing. The Patentee sought to rely on a general passage rather than the detailed paragraph from which the claimed step was drawn. The Board of Appeal refused, finding that "the results of the skilled person's thoughts and reflection do not constitute matter which is a clear and unambiguous consequence of what is explicitly mentioned" (r. 34)

Generalisation success stories

In T 655/24 (Genmab, EP 2943507), the Patentee had better luck with their generalisations. The feature that the protein "does not bind to any Fcγ receptor" had been taken from a page listing it alongside three further conditions about the antibody platform invention. The Board of Appeal found other passages presenting it as one of several alternatives and as particularly relevant, and concluded that the feature was "not inextricably linked with the other features listed on page 14" (r. 32).

Similarly, in T 271/22 (Ceva's EP 2831246) the claim covered a recombinant avian herpes virus with two, three or four inserts, each in its own insertion site. Claim 1 as filed did not require a separate site for each insert, and the examples showed only two inserts. The application as filed, however, stated as an object of the invention that each foreign sequence was inserted in a particular insertion site. This was considered allowable.

In T 722/24, the Opponents attacked a claim to a solid pharmaceutical composition comprising a solid dispersion of enzalutamide on the grounds that only particular solid forms, such as tablets, powders, granules and capsules, had been disclosed, and not solid compositions in general. The Board of Appeal disagreed, finding that solid compositions were disclosed "as a general category", with the specific forms merely "presented as examples" (r. 2.2.2).

In T 792/24, the feature in question was a dose of a compound. The 75 mg/m2 docetaxel dose was the only docetaxel dose mentioned in the application, which described it as the standard dose in the art, and the examples used it alongside other specific features. The Board of Appeal found that the dose was not "inextricably linked" to those features, which merely reflected "standard experimental protocols" (r. 4.3). The added matter objection was thus rejected.

Finally, T 1597/22, relating to a cytisine dosage form, also offers a useful point for functional interrelationship arguments. The Board of Appeal observed that two features contributing to the same effect by different mechanisms are not, for that reason alone, functionally interrelated (r. 2.3.3).

A similarity between these cases is that the Patentees were able to rely on general statements of the relevant features that appeared outside the example. In other words, the saving grace was not a fallback claim or a numbered embodiment but was instead a disclosure in the description saying what the invention was for and that the feature was optional or generally applicable. It, of course, always helps if you are able to identify at filing what embodiments and functional and structural features you eventually want to be able to claim. US-style drafting has a tendency to link embodiments without describing the functional relationship or effect of shared structural features. This is what often creates problems when seeking broader protection in Europe.

Practical guidance

The point that most reliably predicts the outcome of an intermediate generalisation case in both biotech and pharma therefore appears to be whether the application says anywhere, outside the example, what the feature is, what it does, and that it is generally applicable.

Intermediate generalisation is a particular hurdle for biotech and pharma. In this field, we rely heavily on the specific examples and embodiments. The invention also rarely stands still with respect to what the commercial clinical product will end up being and there is a real risk of filing too early (IPKat). Applicants do not always have the luxury of having all the data at the filing date. However, after filing, leads may be optimised for the clinic, doses may change on the clinical trial data, and manufacturing processes and platforms may evolve well beyond those described in the original filing (IPKat). Biotech and pharma inventions can also be highly complex with functionally intertwined features that are difficult to separate. It is often difficult to persuade a Board of Appeal that a feature used in every example has no functional effect and can simply be left out of the claim.

The obvious answer to this problem might seem to be to generate data showing which features are not essential. However, this is unfortunately usually not remotely feasible. Demonstrating that each component can be varied or removed would require controls for every element of the product. For a CAR-T construct, this might mean testing alternative hinges, transmembrane domains and costimulatory domains. For an LNP, it might mean varying each lipid and the molar ratios between them. For a formulation, it might mean removing or substituting each excipient and buffer in turn. For a cell therapy process, it might mean testing each growth factor, time point and culture step. Running all of these controls would be prohibitively expensive, and it is understandably not a priority for the stretched R&D departments of most biotechs.

The only alternative is a drafting style that is appropriate for supporting a broad scope of claims (IPKat). Statements that a feature is optional, that it is one of several alternatives, or that a component can be used independently of the others are what have rescued the Patentees in the success stories above. The challenge is always how to include such statements without straying into the patent profanities that may cause trouble in the US (IPKat). This is a difficult line to tread. Applicants also need to decide at filing how broadly they want to be able to claim, or whether they are content to be limited to the features of their clinical lead.

Final thoughts

To add to the complexity on this issue, we now await the outcome of the referral in G1/26 (referred from T 873/24, IPKat). The third question of this referral asks whether a claim term must be assessed against every interpretation that makes technical sense on the claim alone, or only against those established from the specification as a whole. Whilst this is not a case directly related to intermediate generalisations, it may nonetheless influence the construction step that every question of added matter begins with (IPKat). The outcome in G1/26 may therefore have an interesting impact on how broadly a Board of Appeal reads a claim it is testing for added matter and intermediate generalisations.

Correctly navigating this line is becoming ever more important in the biotech and pharma field. Broad platform technologies are increasingly attractive in cell and gene therapy and next-generation biologics, and claims broad enough to catch fast-followers are correspondingly more valuable (IPKat). Being able to generalise beyond a specific sequence, a specific construct or a specific embodiment of the product may make the difference between a valuable patent and one that is easily designed around by a fast-follower that quickly erodes your market (IPKat).

Further reading
Do you want to reuse the IPKat content? Please refer to our Policies section. If you have any queries or requests for permission, please get in touch with the IPKat team.
Print | Share by email

Comments

  1. I am not a specialist in biotech and have great respect for people who are able to deal with such cases in a manner like Dr. Hughes does.

    A general comment: when it comes to added matter, adding “may” in nearly every sentence of the description does not help to derive directly and unambiguously any combination of features. I have seen this in so many applications of US origin. It has the same effect as some boilerplate statements that any feature can be combined with any feature. See T 2241/15, Reasons 4.8, last sentence, or T 1203/13, Reasons 2, penultimate §. If everything is optional, nothing is optional.

    In T 350/20, the board held that too many options in a forest of optional features can lead to the infringement of Art 123(2).

    The “Essentiality test” is not to the liking of the boards, but is still present in the Guidelines H-V, 3.1, under the name “Three point test”. Applied correctly it should lead to the same conclusion than the “gold standard”, i.e. “directly and unambiguously derivable”. My experience, at least in other areas than biotech, shows that it can work quite well, but without the use of the forbidden word “essential”.

    Functional features allow a broad scope of protection. One example is in principle sufficient, cf. Guidelines F-IV, 6.5, but it appears always better to give more than one example in order to show that a broad functional feature is allowable. In T 321/23, the proprietor requested the board to refer questions to the EBA on intermediate generalisations, but this request was refused.

    ReplyDelete
  2. In the present time one might find that with most specifications if the amendment is not the same as the language of a dependent PCT claim, then the applicant has an uphill struggle to convince an Examiner added matter is not present in the resulting combination. This of course protects third parties from having to predict the millions of possible claim amendments that are possible when amending based on the description. In addition, it only seems fair that one cannot used the Examples for basis for an amendment. That was not their intended purpose when the specification was drafted, and why should parties have to go through the uncertainty of what sort of features could be extracted from them for claim amendments?

    ReplyDelete

Post a Comment

All comments must be moderated by a member of the IPKat team before they appear on the blog. Comments will not be allowed if the contravene the IPKat policy that readers' comments should not be obscene or defamatory; they should not consist of ad hominem attacks on members of the blog team or other comment-posters and they should make a constructive contribution to the discussion of the post on which they purport to comment.

It is also the IPKat policy that comments should not be made completely anonymously, and users should use a consistent name or pseudonym (which should not itself be defamatory or obscene, or that of another real person), either in the "identity" field, or at the beginning of the comment. Current practice is to, however, allow a limited number of comments that contravene this policy, provided that the comment has a high degree of relevance and the comment chain does not become too difficult to follow.

Learn more here: http://ipkitten.blogspot.com/p/want-to-complain.html

← Previous Post Next Post →