Generics (UK) Ltd & Ors v AstraZeneca AB  ([2025] EWCA Civ 903)
Court of Appeal Confirms the “Ab Initio Plausibility” Standard and Requires a Distinct Technical Contribution for Selections from Prior Art

1. Introduction

On 16 July 2025 the Court of Appeal (Civil Division) handed down a judgment of major importance for pharmaceutical patent litigation. The appeal, brought by AstraZeneca AB against three generic manufacturers (Generics (UK) / Teva / Glenmark), concerned the validity of a patent covering the blockbuster drug dapagliflozin (Forxiga) and its corresponding Supplementary Protection Certificates (SPCs).

At first instance Dr Michael Tappin KC invalidated the patent for (i) lack of inventive step and insufficiency on the ground of plausibility, and (ii) on the further ground that the patent was merely an arbitrary selection from an earlier broad disclosure (WO 01/27128). AstraZeneca appealed on eight grounds.

The Court of Appeal (Arnold LJ giving the leading judgment, with Peter Jackson LJ and Stuart-Smith LJ concurring) dismissed the appeal in full. In doing so, the Court:

• Re-affirmed the “ab initio plausibility” test laid down by the Supreme Court in Warner-Lambert for both sufficiency and inventive-step challenges involving chemical compounds.
• Clarified that the Enlarged Board of Appeal’s recent decision G 2/21 does not dilute that standard in the UK, nor justify departing from the Court of Appeal authority in Sandoz v BMS.
• Re-stated the law on arbitrary selections from a disclosed genus: a patentee must show the selected compound provides a genuine technical advance, not merely the same advantage more narrowly.

2. Summary of the Judgment

The Court upheld the trial judge’s findings that:

• The patent did not make it plausible, on its face and read with common general knowledge, that dapagliflozin is an SGLT2 inhibitor potent enough to treat diabetes.
• References in the patent to an assay were bare methodological descriptions; they did not amount to a “verbal statement of positive results”.
• When cross-referenced documents (WO 128 and four Tanabe papers) were considered—as the skilled team would do—they actually generated doubt about dapagliflozin’s activity.
• Because WO 128 already asserted SGLT2 inhibition for a vast class that includes dapagliflozin, and because the patent did not show any special advantage of dapagliflozin over close neighbours (e.g., Example 12 in WO 128), the claim was merely an arbitrary selection and thus obvious/lacking inventive step.

Consequently both the compound claim (claim 2) and the medical-use claim (claim 15) were invalid for insufficiency and obviousness; the SPCs therefore fell with the patent.

3. Analysis

3.1 Precedents Cited and Their Influence

• Warner-Lambert v Generics (UK) [2018] UKSC 56 – Supreme Court majority required that the patent make the therapeutic effect “plausible” at the filing date. Central authority binding the Court.
• Sandoz v BMS [2023] EWCA Civ 472 – Arnold LJ extended Warner-Lambert to single-compound product claims; same standard for inventive step and sufficiency. Key authority challenged by AstraZeneca.
• G 2/21 (EPO Enlarged Board, 2023) – considered post-published evidence and formulated a two-limb test (effect “encompassed by” and “embodied by” the original technical teaching). AstraZeneca argued this diluted the UK plausibility threshold.
• Dr Reddy’s v Eli Lilly [2010] RPC 9 and subsequent CA cases – established the “arbitrary selection” doctrine: a mere pick from prior art needs an additional technical effect.

The Court reviewed these authorities in detail and concluded:

“Neither G 2/21 nor foreign jurisprudence establishes a settled view contrary to Sandoz v BMS. Until the position clarifies, UK courts must continue to apply ab initio plausibility.” (Arnold LJ §128)

3.2 Legal Reasoning

3.2.1 Plausibility Findings

1. Identify Claim Scope: Claim 2 – dapagliflozin per se; Claim 15 – use of dapagliflozin to treat diabetes.
2. Identify Technical Effect: For both claims the relevant effect is treating diabetes by selective SGLT2 inhibition.
3. Assess Disclosure: The patent included:
   • Repeated assertions that formula I “is an SGLT2 inhibitor”.
   • An assay description but no data and no explicit reference that dapagliflozin itself had been tested.
   The Court held this was a “bare assertion”, not even a verbal statement of results.
4. Common General Knowledge & Cross-References:
   • WO 128 (same wording) contained no data.
   • Tanabe papers suggested phlorizin analogues required particular structural features missing in dapagliflozin, creating doubt.
5. Conclusion on Plausibility: A skilled reader—armed with the CGK and conscious of the gaps—would not consider it plausible (let alone beyond legitimate doubt) that dapagliflozin possesses the claimed therapeutic potency.

3.2.2 Inventive Step vs Sufficiency Post-G 2/21

AstraZeneca urged the Court to adopt a two-tier approach:

• For pure compound claims (inventive step) — apply “ab initio implausibility” (i.e., allow unless there is positive reason to doubt).
• For medical-use claims (sufficiency) — keep “ab initio plausibility”.

The Court rejected the invitation because:

• Sandoz v BMS settled that the same standard applies; Actavis exception for following EPO case law did not bite because EPO jurisprudence is not yet “settled”.
• Text and context of G 2/21 do not mandate or even clearly support a lower threshold.
• Allowing speculative filings based on absence of reason to doubt would undermine the first-to-file system and the patent bargain (see §§11–15 of judgment).

3.2.3 Arbitrary Selection

Key questions:

1. Is dapagliflozin merely one member of a disclosed genus (Formula IB) in WO 128? – Yes.
2. Does the patent teach any distinct technical advantage over Example 12 or the genus? – No (AstraZeneca expressly disavowed superiority).
3. Does making the same property “plausible” for the single compound count as technical contribution? – No. The contribution must be the property itself, not the plausibility.

Therefore the selection was “arbitrary”, rendering claim 2 obvious and claim 15 likewise because its therapeutic assertion rested on the same unsubstantiated premise.

3.3 Potential Impact of the Decision

• Plausibility remains stringent: UK litigants must still satisfy the ab initio plausibility standard for product and medical-use claims notwithstanding G 2/21.
• Drafting consequences: Applicants should include at least indicative data or robust mechanistic reasoning for each claimed compound. Omitting such support risks fatal attack even after clinical success.
• Cross-referenced documents: Skilled readers may look outside the patent when it cites other literature; those documents may increase doubt, so patentees must draft with caution.
• Arbitrary selection clarified: Simply narrowing a claim to one compound and re-stating the same advantage is not inventive; a specific, demonstrated improvement is required.
• UK–EPO interplay: Until the Boards of Appeal converge on a single post-G 2/21 approach, UK courts will not depart from their own precedent; users must manage the risk of divergent outcomes across Europe.
• SPC vulnerability: When the basic patent is knocked out for plausibility/arbitrary selection, any SPC anchored to it will fall – critical for life-cycle management strategies.

4. Complex Concepts Simplified

• Plausibility: A commonsense check: does the patent give the skilled person some rational basis to expect the promised benefit at the filing date? Not proof, but more than wishful thinking.
• Ab Initio Plausibility vs. Ab Initio Implausibility:
  • Plausibility – patent must affirmatively make effect credible.
  • Implausibility – effect stands unless the patent gives reason to doubt. UK remains with the former.
• Arbitrary Selection: Picking one item out of a disclosed shopping list is not inventive unless you show why that pick is special (e.g., better potency, safety, stability).
• Technical Contribution: The “price” the patentee pays for exclusivity – a genuine advance in knowledge or capability supplied to the public through the patent disclosure.
• Post-Published Evidence: Data generated after filing can support a patent only if the effect was already plausible; it cannot retrofit an invention into the application.

5. Conclusion

Generics (UK) v AstraZeneca cements two pillars of UK patent law:

1. The ab initio plausibility test, as articulated in Warner-Lambert and confirmed in Sandoz v BMS, remains the yardstick for both sufficiency and inventive-step assessments of chemical and medical inventions, notwithstanding G 2/21.
2. A selection from broad prior art is inventive only if the selected compound brings a demonstrable, plausible technical benefit over its peers; merely isolating and re-asserting the same property is inadequate.

By declining to dilute these standards, the Court protects the integrity of the patent system, preserving the balance between incentivising innovation and preventing unwarranted monopolies. Practitioners should heed the judgment’s drafting lessons: include meaningful data or cogent scientific rationale at filing, and be wary of relying on undisclosed “secret” results. Until the EPO’s post-G 2/21 jurisprudence stabilises, divergence across Europe will persist, but in the UK the message is unmistakable — plausibility is king, and arbitrary selections will be scrutinised for real technical contribution.