UPC upholds pharmaceutical patent for subgroup of patients: Key lessons for patent holders and challengers

On 9 October 2026, the Milan section of the Central Division of the Unified Patent Court (UPC) rejected Biogaran’s revocation action against Lundbeck’s European patent EP 4 223 296 (UPC_CFI_1587/2025). The patent, which has unitary effect, covers nalmefene for reducing alcohol consumption in a narrowly defined group of alcohol-dependent patients. The court maintained the patent in amended form, following Lundbeck’s unconditional deletion of two dependent claims.

The decision matters beyond the pharmaceutical sector. It shows how the UPC treats second medical use claims built on a patient subgroup, how strictly it applies its front-loaded procedure, and how it coordinates with parallel opposition proceedings before the European Patent Office (EPO). In a single action, a successful challenger could have removed protection in every UPC Contracting Member State. The rejection of the action leaves the patent in force, in its amended form, across the participating territory.

The dispute

EP 4 223 296 is the last link in a chain of divisional applications derived from WO 2014/001427 (WO 2014/001427 → EP 2 866 808 → EP 3 138 564 → EP 3 345 604 → EP 4 223 296). Two earlier family members have already been revoked: EP 2 866 808, at the proprietor’s own request, and EP 3 138 564, by the Technical Boards of Appeal (T 963/20). The patent was granted on 29 January 2025 and its unitary effect was registered on 5 February 2025.

Biogaran, the biggest generic manufacturer in France, filed its action in Milan on 14 November 2025, seeking revocation for all UPC Contracting Member States on grounds of added matter, insufficiency, lack of novelty, and lack of inventive step. An EPO opposition is running in parallel. In March 2026, Lundbeck sought provisional measures against Biogaran before the Paris Local Division, relying on both EP 3 345 604 and EP 4 223 296. After the court proposed hearing the two cases in parallel, the parties agreed that the application for provisional measures would be withdrawn.

Procedure

Biogaran’s rejoinder introduced five new documents and fresh priority, novelty, and inventive step attacks, partly prompted by the EPO Opposition Division’s preliminary opinion of 6 May 2026. The court refused to consider them, as Biogaran had not sought leave under Rule 263 of the Rules of Procedure (RoP) to amend its case, the documents were already available from the start, and UPC proceedings are front-loaded. Relying on its earlier order in EOflow v. Insulet, the court recalled that a claimant must present its full case, including with respect to prior art, in its statement for revocation, and that late filing of available prior art deprives the other side of a full adversarial debate (Article 76 UPCA). A preliminary opinion from the EPO proceedings therefore cannot serve as a back door.

By contrast, the court admitted Lundbeck’s late, unconditional request to delete dependent claims 7 (treatment period) and 9 (patient age) under Rule 30.2 RoP. Applying the Court of Appeal’s test in Meril v. Edwards (a later amendment may be admitted where it could not reasonably have been made earlier and does not unreasonably hinder the other party), the court noted that the deletion mirrored an amendment filed before the EPO after the Opposition Division’s preliminary opinion, added no technical matter, and simplified the dispute. The court treated the limitation made in the parallel EPO proceedings as a fait accompli and considered it necessary to examine a consistent claim text in both fora. The court also highlighted the proprietor’s duty under Rule 30.3 RoP to inform it of parallel proceedings.

What the claim covers

The court read claim 1 as a claim directed to a patient subgroup rather than a second-medical-use claim. The patient must be alcohol dependent, must have a high drinking risk level (mean daily consumption exceeding 60 g of pure alcohol for men or 40 g for women), and must still show that level after a one-to-two-week observation period, which excludes “early reducers.” Nalmefene is then administered on an as-needed basis, on days when the patient perceives a risk of drinking, preferably one to two hours beforehand. The skilled person is a clinician experienced in treating patients who may be alcohol dependent.

Priority also mattered. The first priority was conceded to be not validly claimed, but the claim remained entitled to the second priority document (2 November 2012). The Summary of Product Characteristics for the related medicinal product SELINCRO®, a 2013 paper by Mann, and the EMA assessment report of December 2012, were therefore not prior art.

No impermissible added matter and no insufficiency of disclosure

Impermissible added matter arises where a granted claim contains subject matter extending beyond the content of the application as filed (Article 138(1)(c) EPC, applied through Article 65(2) UPCA).

The court applied the “directly and unambiguously derivable” test previously set out by the Court of Appeal on 13 August 2026. It stressed pleading discipline: A challenger must compare the granted claim with the original disclosure and identify what was extended. The fact that the parent application contained 135 embodiments and, in theory, millions of combinations, thus does not suffice: A claim may raise clarity or drafting issues, but not added matter, and the court cannot cure a vague objection of its own motion without compromising its impartiality. Given uncertainty over the required specificity, pending further guidance from the Court of Appeal, the court nonetheless checked the application as filed, acknowledging a limited ex officio assessment, and found a direct basis in the cross-referenced embodiments E24 to E26, whose “and/or” wording expressly includes the claimed combination, with continuity of disclosure throughout the divisional chain.

On the alleged insufficiency of disclosure (Article 83 EPC), the court followed the Court of Appeal’s test in Amgen v. Sanofi: The patent, read as a whole, must enable the skilled person to carry out the invention without undue burden, and the claimed therapeutic effect must also be credible across the relevant claim scope. Biogaran’s three attacks failed: The trials excluded certain patient groups, motivational counselling was not claimed, and the trials used only an 18 mg dose while claim 1 recites none. Trial exclusion criteria do not make the effect implausible without verifiable facts, and the burden lies on the challenger. Counselling (the BRENDA model) was an adjunct, not an essential feature. On dose, the court found persuasive, though not binding, the EPO case law (G 2/21, T 728/21) requiring a second medical use to be credible at the filing date, but held that credibility may rest on the patent, the prior art, and common general knowledge. Karhuvaara already showed efficacy at 10 mg, dependent claims recite 10 – 20 mg, and selecting a safe and effective dose is routine.

Novelty and inventive step

For a known drug, defining a patient subgroup does not itself establish novelty: The subgroup must be physiologically or pathologically distinct and functionally linked to a distinct therapeutic effect. The patent’s phase III data showed a more pronounced effect in patients with a high drinking risk level at baseline who maintained it after the observation period, which the court treated as a genuine technical teaching. A key prior art starting point, the Karhuvaara study, enrolled heavy drinkers rather than patients selected for dependence (about 7% did not meet DSM-IV criteria), had no observation period, and performed no subgroup analysis.

On inventive step, the court applied the Court of Appeal’s November 2025 approach (Meril v. Edwards; Amgen v. Sanofi). It framed the objective problem as identifying a subgroup of alcohol-dependent patients who benefit more from nalmefene, without pointers to the claimed solution, and accepted Karhuvaara and Lundbeck’s 2011 press release as realistic starting points. Neither suggested selecting patients by drinking risk level after an observation period. The other documents did not fill the gap: some were tentative or silent on stratification, and run-in studies from cardiology and cocaine dependence gave no reason to be consulted. Drawing on its own May 2026 ruling (UPC_CFI_552/2025), the court distinguished hope of success (what the skilled person could do) from reasonable expectation of success (what they would do), and noted that the patent’s own data, showing over 50% early reducers, sat poorly with the idea that excluding them was routine.

An obviousness case that can be built only by combining more than 10 documents on different technical questions, none supplying the missing pointer, reveals hindsight. Finally, the EPO Technical Boards of Appeal’s revocation of parent EP 3 138 564 (T 963/20) had no res judicata effect: The two patents are legally distinct, and the EPO and the UPC belong to different legal systems, so an EPO decision may persuade but does not bind the UPC.

Key takeaways

Several practical lessons stand out.

First, patient selection claims can be defended on the clinical data specified in the description, and an attack on credibility needs verifiable facts, not general doubts about trial design.

Second, omitting a precise dose from an independent second-medical-use claim does not necessarily entail insufficiency, provided the patent, prior art, and common general knowledge enable the skilled person to select a safe and effective dose without undue burden. Similarly, impermissible added matter requires a strict assessment that technical information could not be derived directly and unambiguously from the application as filed.

Third, and most relevant for generic challengers, the front-loaded procedure leaves little room to repair a (too) thin statement for revocation or counterclaim for revocation: All available prior art and attacks must be gathered before filing, and a favorable EPO preliminary opinion cannot be relied upon as a back door.

Fourth, for patentees, a limitation aligned with parallel EPO proceedings may be admitted late if it narrows the dispute, provided the court is kept informed of those proceedings.

Finally, an adverse EPO ruling on a related family member has no automatic binding effect in UPC proceedings concerning a legally distinct patent.

Interestingly, the court also ordered Biogaran to bear Lundbeck’s costs and pay an interim costs award of €100,000 within three weeks.