Generics (UK) v AstraZeneca (EWCA, [2025] EWCA Civ 903, 2025)

Dhenne Avocats · 11 October 2026

Our analysis

Summary

The Court of Appeal upheld the invalidity of AstraZeneca’s European Patent (UK) No. 1 506 211 for dapagliflozin, and hence of the SPCs based on it, for lack of inventive step and insufficiency (paras 2 and 144). The patent did not make it plausible that the compound would be useful to treat diabetes: the assay description did not say that dapagliflozin had been tested, still less with what activity (paras 85, 86 and 90). The court declined to depart from Sandoz v BMS in the light of G 2/21, as neither the Boards of Appeal nor national courts had reached a settled view (paras 108, 115 and 128), and also upheld the finding of arbitrary selection from the genus disclosed in WO 01/27128 (paras 140 and 144). The judgment confirms that, in English law, the technical effect of a compound must be made plausible by the application itself.

Facts and procedure

The patent expired on 14 May 2023 and underpins SPC/GB13/021 and SPC/GB14/050, expiring on 13 and 14 May 2028 (para 2). Generics (UK), Teva and Glenmark contended that the patent did not make plausible that dapagliflozin was an SGLT2 inhibitor or useful to treat diabetes, and that it was an arbitrary selection from the compounds of Bristol-Myers Squibb’s WO 128, defined by a very broad Markush formula (paras 4 and 5). Dapagliflozin differs from the compound of Example 12 of WO 128 only in having an ethoxy rather than a methoxy group (para 63). On 28 April 2025 the Patents Court (Michael Tappin KC) held the patent invalid on both grounds ([2025] EWHC 1012 (Pat)) (para 7). On appeal AstraZeneca relied on claim 2 (the compound) and claim 15 (its use in the manufacture of a medicament for diabetes) (para 59).

The applicable law

Articles 56 and 83 EPC, implemented by the Patents Act 1977, do not mention plausibility, which is a creature of case law (para 9). Under Warner-Lambert, the application read with the common general knowledge must make the claimed therapeutic effect plausible (paras 16 and 17). Sandoz v BMS applies that standard to inventive step and sufficiency for a claim to a single compound (para 19). A selection from the prior art is inventive only if it makes a technical contribution (Dr Reddy’s) (paras 38 to 40).

Question

Did the patent make the utility of dapagliflozin plausible, did G 2/21 require the less demanding ab initio implausibility standard for the product claim, and was dapagliflozin an arbitrary selection?

Decision

The assay passage describes a method without identifying the « inhibitor » tested or stating any result (paras 85 and 86); it gives no reason to think that dapagliflozin was tested (para 88) and, even if it had been, the patent says nothing of its potency or EC50, the repeated statements of utility being pure assertion (para 90). Warner-Lambert binds the court on the sufficiency of medical use claims (para 97), and claim 2, broader than claim 15, cannot be subject to a less stringent test (para 94). G 2/21 does not justify departing from Sandoz v BMS: the distinction between the two standards is inescapable (paras 106 and 108), the Boards of Appeal have not adopted a settled view, as T 116/18 and T 314/20 show (paras 109 to 115), and courts in other Contracting States are some way from unanimity (para 127). Even on a legitimate reason to doubt test, the Tanabe Seiyaku papers cited in WO 128 would give the skilled team reason to doubt (para 133). Finally, mere plausibility cannot make a selection inventive: the compound must in fact have a useful property (para 140), and nothing distinguishes dapagliflozin from the compound of Example 12 (paras 143 and 144). The appeal was dismissed (para 144).

Key points for practice

• Describing an assay without results or without identifying the compound tested does not make the utility of the claimed compound plausible (paras 85, 86 and 90).
• For the Court of Appeal, G 2/21 does not displace the ab initio plausibility standard, absent settled Board of Appeal case law (paras 108 and 115).
• A selection from a prior genus requires a useful property that sets the compound apart; confirming what the prior art already asserted is not enough (paras 140 and 143).
• Practical point: data kept secret at filing cannot be relied on; the application must contain results or reasoning making the effect plausible (paras 13, 119 and 145).

Provisions applied

European Patent Convention

Art. 52; Art. 56; Art. 83; Art. 138

National law

Patents Act 1977, sections 1(1)(a), 3, 14(3) and 72(1)

Case law cited

Generics (UK) v Warner-Lambert [2018] UKSC 56; Sandoz v Bristol-Myers Squibb [2023] EWCA Civ 472; EPO, G 2/21; EPO, T 116/18; EPO, T 314/20; Dr Reddy’s v Eli Lilly [2009] EWCA Civ 1362; Generics (UK) v Yeda [2013] EWCA Civ 925; Akebia v FibroGen [2021] EWCA Civ 1279; Actavis v Merck [2008] EWCA Civ 444; Human Genome Sciences v Eli Lilly [2011] UKSC 51; Court of Appeal of The Hague, ECLI:NL:GHDHA:2023:1593; Swiss Federal Patent Court, Mepha v Bristol-Myers Squibb, 5 March 2024

Related decisions

• Sandoz v Bristol-Myers Squibb, 4 May 2023, the precedent applied, from which the court declined to depart.
• AstraZeneca v Glenmark, 16 April 2025, interim injunction in the same dapagliflozin dispute.
• Bristol-Myers Squibb v Sandoz, 15 August 2023, Dutch apixaban ruling discussed by the court.
• Mepha v Bristol-Myers Squibb, 5 March 2024, Swiss ruling on G 2/21 discussed by the court.

Prepared by Dhenne Avocats from the text of the decision (National Archives, Find Case Law), consulted on 11 October 2026. Only the official text is authoritative.