UPC_CoA_19/2026 – Sophia v Guardant

Expert Commentary

No urgency Facts 1. The Paris Local Division refused to grant a preliminary injunction (PI) sought by Guardan against Sophia on the basis of EP 073, EP 066 and EP 986, because it held that it was more likely than not that the patent at issue on appeal (EP 066) was invalid due to added matter. 2. Guardant was ordered to pay interim costs € 400.000. 3. Guardant appealed only with respect to EP 066. Sophia filed a cross-appeal. The Court of Appeal 1. International jurisdiction: Sophia contests the Court’s jurisdiction with respect to the infringements in Spain and Switzerland. 2. The Court refers to Solvay v Honeywell and accepts jurisdiction on the basis of Art. 8 sub 1 of the Brussels Regulation, holding that both the anchor defendant and Sophia Switzerland are jointly responsible for the alleged infringements in Spain and Switzerland. 3. The Court describes the invention and formulates as object “to provide as improved method for using cfDNA to detect and monitor cancer”. 4. The Court defines the skilled person, interprets the claim, and disagrees with Sophia’s argument that the claim requires the identification of the cancers. 5. The Court also states that it is sufficient for the claimed method to be used for the detection of only one of the four cancers mentioned in the claim. 6. With respect to feature 1.6, the Court accepts Sophias interpretation. 7. New legal arguments can be raised on appeal. Claim interpretation is a matter of law. 8. The Court agrees with Guardant that the Paris Local Division was wrong in holding that the patent was more likely than not to be invalid because of added matter. Implicitly disclosed matter shall also be considered as part of the content of a patent application. 9. Also here the Court states that added matter is a question of law and the Court is entitled to consider the entire document. Accordingly, Guardant was allowed for the first time to rely on example 2 during the oral hearing at first instance. 10. Urgency: the Court finds that the facts alleged and the evidence submitted by Guardant are insufficient to conclude that Guardant acted without undue delay. 11. If a patentee owns several patents but already has all the evidence of infringement of one of them and nevertheless fails to act, this may constitute unreasonable delay. 12. A patentee does not have to start proceedings before having sufficient evidence. Hence, it is for the patentee who files for a PI to explain, in a substantiated manner, what further analysis was required, why this was necessary, how long it took, and why that period was reasonable. 13. The Court concludes that Guardant failed to establish urgency. 14. A cross-appeal is inadmissible where it challenges only the reasoning of the decision and not its operative part. 15. The same applies costs. The Court of Appeal decides on its own about the costs of both the appeal and the first instance. Therefore, if the Court in first instance did not make a decision on costs, the Court of Appeal may do so itself and therefor a cross appeal is not admissible. Comment 1. The requirement of urgency for a PI should be taken seriously. With respect to knowledge of the infringement, a patentee cannot rely on formalistic arguments (“the patent was not mentioned in the warning letter”, etc.). Although it is not required to actively monitor the market, it should be considered to have knowledge if, in view of the circumstances, a reasonable person would have become aware of the alleged infringement. 2. A patentee does not have to start proceedings before it has its case fully prepared, but it should be able to show, in reasonable and convincing manner, that it acted diligently, with a sense of urgency. The problem is that in big companies the decision-making is often slow and that internal labs responsible for testing are slow too. Many people nowadays work only 4 days a week, and take 5 weeks’ vacation (indeed, in this case the delay appears to have occurred during the vacation period). So this is a wake up call: urgent is really urgent. If a matter is truly urgent, it cannot wait. So if a patentee waits too long the matter is not urgent! I think in the UPC context where you can get a decision in around 12 to 14 months, such a rather strict approach to urgency is justified. 3. We know that claim interpretation, added matter (and, of course, inventive step) are questions of law. A party may therefore have a different interpretation of a claim on appeal (or even during in the oral hearing at first instance), provided that the reasoning is based on the same facts. However, how does that work in practice? In this case, the Court did not make a problem (because added matter is a legal question) that the patentee for the first time referred to example 2 during the oral hearing to argue that there was no added matter. But is that not, in reality, a new argument, or perhaps even a new fact? Anyway, I do not find this entirely satisfactory. The other party can be taken by surprise when, after the whole discussion between the parties has concentrated on completely different arguments, that party all of a sudden relies on example 2, which has never been discussed in the context of added matter. I accept that, where an issue is one of law, the Court may, even if parties have not referred to it, conclude for its own reasons that there is added matter by relying for instance on example 2. However, would that not qualify as a surprise decision? If so, should the parties not be given an opportunity to comment on it?

Full Decision Text

Court of Appeal of the Unified Patent Court UPC-CoA-19/2026 Final Order 2 July 2026 Patent: EP 3 443 066 **Decision** of the Court of Appeal of the Unified Patent Court issued on 2 July 2026 concerning a request for provisional measures **HEADNOTES:** 1. A patent holder is not obliged to assert all patents in one application for provisional measures. If a patent holder has the necessary information for filing an application for provisional measures regarding some, but not all patents, delaying the filing of the Application for provisional measures until it has the necessary information regarding all allegedly infringed patents may be regarded as an unreasonable delay. 2. If a patent holder is aware, on the basis of a document, that one or more of its patents has been infringed, it must not turn a blind eye to the fact that the document also indicates the infringement of one or more of its other patents. 3. As a general rule, the patent holder is not obliged to monitor the market. However, if the patent holder becomes aware of specific circumstances that suggest patent infringement, the patent holder is expected to investigate the matter with due diligence so that the patent holder can take action against all infringers as quickly as possible. 4. A long-standing infringement may indicate that the patent holder has turned a blind eye to the infringement. 5. If the applicant is a legal entity, the decisive factor for assessment of urgency is when its authorised representative body or an individual, who by virtue of its position within the legal entity is capable of pursuing the infringement or reporting it, internally becomes aware of it. For example, this person could be an employee of the legal department, a senior member of the sales department or a mystery shopper engaged specifically to investigate intellectual property rights infringements. 6. As a general rule, the patent holder may wait a reasonable time for a response to a warning letter before drafting and lodging an application for provisional measures. 7. As a general rule, the unsuccessful party must bear the reasonable and proportionate legal costs and other expenses incurred by the successful party (Art. 69(1) UPCA). Exceptions apply where equity requires a different allocation of costs, in particular where a party succeeds only in part or in exceptional circumstances (Art. 69(2) UPCA) and where the party has caused unnecessary costs to the Court or another party (Art. 69(3) UPCA). As a rule, the withdrawal of an action or application means that the defendant must be considered the successful party. Equity does not provide otherwise in cases where the claimant applies to withdraw the application for provisional measures or the action after the defendant, for the first time in the proceedings and not in response of the warning letter, raises an argument against infringement. In principle, a person who has received an unjustified warning is under no obligation to provide clarification why it will not comply with the request. **KEYWORDS:** - Claim interpretation, relevance of embodiments - Added matter (Article 138(1)(c) EPC) - Urgency (R. 211.4 RoP): multiple patents, lack of substantiation, general no obligation to monitor the market - Cost allocation and equity (Article 69 UPCA): response to warning letter - Missing cost allocation/cross-Appeal **APPLICANT AND APPELLANT (APPLICANT BEFORE THE COURT OF FIRST INSTANCE )** Guardant Health, Inc., Palo Alto, California, United States of America (hereinafter: “Guardant”) represented by attorney at law Agathe Michel-De Cazotte, Carpmaels & Ransford LLP, London, United Kingdom **RESPONDENTS ( AND DEFENDANTS BEFORE THE COURT OF FIRST INSTANCE )** 1. Sophia Genetics SA, Rolle, Switzerland 2. Sophia Genetics SAS, Bidart, France 3. Sophia Genetics SRL, Milan, Italy 4. Sophia Genetics GmbH, Freiburg, Germany (hereinafter for all: “Sophia” or “the Sophia companies”) represented by attorney at law Liz Cohen and other representatives from Bristows (Ireland) LLP, Dublin, Ireland **PATENT AT ISSUE** EP 3 443 066 **PANEL AND DECIDING JUDGES** Panel 2 This order has been issued by Rian Kalden, presiding judge and legally qualified judge Patricia Rombach, judge-rapporteur and legally qualified judge Ingeborg Simonsson, legally qualified judge Arwed Burrichter, technically qualified judge Andreas Gustafsson, technically qualified judge **IMPUGNED ORDER OF THE COURT OF FIRST INSTANCE** Paris Local Division, 23 January 2026, in the provisional measures proceedings UPC CFI 808/2025 **ORAL HEARING** 18 May 2026 **LANGUAGE OF THE PROCEEDINGS** English **FACTS** 1. Guardant is the proprietor of the European Patent 3 443 066 with unitary effect (patent at issue). The application was published on 20 February 2019 and the mention of the grant of the patent was published on 2 October 2024. The patent at issue claims priority of 14 April 2016 and 18 April 2016 based on six US applications. No opposition was filed against the patent. In addition to its UPC designation it is in force in Switzerland, Spain and UK. 2. Claim 1 of the patent at issue reads as follows: “A method for detecting the presence or absence of colorectal cancer, ovarian cancer, lung cancer or pancreatic cancer in a subject comprising: sequencing circulating cfDNA from the subject at a depth of at least 50,000 reads per base to detect one or more genetic variants associated with cancer, Wherein the sequencing is performed on an enriched set of amplified cfDNA molecules which comprises a panel of genomic regions, wherein the genomic regions in the panel comprise one or more loci from each of the genes AKT1, ALK, APC, ATM, BRAF, CTNNB1, EGFR, ERBB2, ESR1, FGFR2, GATA3, GNAS, IDH1, IDH2, KIT, KRAS, MET, NRAS, PDGFRA, PIK3CA, PTEN, RB1, SMAD4, STK11 and TP53, and further comprising amplifying the cfDNA prior to sequencing, and determining a consensus sequence from sequence reads obtained from the sequencing to reduce errors from amplification or sequencing.” 3. Guardant markets liquid biopsy tests such as “Guardant 360”. 4. The Sophia companies are part of the Sophia Genetics group (hereinafter “Sophia Genetics”). Respondent 1 is the parent company based in Switzerland, Respondents 2 to 4 are subsidiary companies based respectively in France, Italy and Germany. Sophia Genetics is a cloud-native healthcare technology group and creator of SOPHIA DDMTM, a platform that analyses genomic and multimodal data and generates real-time, actionable insights for hospitals, laboratories, and biopharma institutions. 5. Sophia offers and supplies the “MSK-ACCESS® powered with SOPHIA DDMTM” (hereinafter “attacked embodiment” or “MSK-DDM”) in the UPC territories and inter alia in Spain, Switzerland and Norway. 6. On 27 May 2025 Guardant sent a letter (Exhibit SG49, hereinafter: “UK warning letter”) to Sophia Genetics SA (Respondent 1) and Sophia Genetics Ltd. (not party in this proceedings), explaining its view that the MSK-DDM test falls within the scope of various UK patents, including the UK designations of EP 3470533 (“EP‘533”) and EP 3591073 (“EP‘073”). As far as it is relevant here it has the following content: “Guardant has, from publicly available sources, become aware of the details of: a) Sophia’s liquid biopsy test ‘MSK- ACCESS® powered with SOPHIA DDMTM’ for which details are disclosed on your website (...) in the enclosed ‘factsheet’ which is provided via your website (your document ‘GL-GN-2400004-r5’), and in the user manual (...), and b) Sophia’s offer to supply of the ‘CUMIN Adapters’ product for use with your liquid biopsy test e.g. as outlined in the enclosed certificate of analysis. Based on this publicly available information, Guardant’s position is that Sophia’s ‘MSK-ACCESS® powered with SOPHIA DDMTM’ test falls within the scope of protection conferred by each of the Patents. (...) To avoid unnecessary litigation (...) please confirm as soon as possible, and in any event by 6 June 2025, that Sophia and/or any affiliates thereof (...) will: (a) respect Guardant’s patent rights; and thus (b) not seek to market or offer for use... Guardant reserves all its rights in relation to this matter, including the right to issue proceedings and seek urgent preliminary injunctive relief, if necessary, without further notice.” 7. Sophia replied on 20 June 2025 to the UK warning letter. In the UK, Guardant started litigation on 14 July 2025. In the UK action, the UK designations of EP‘533, EP 3766986 (EP‘986) and the patent at issue are asserted as well as a national patent in the same family as EP’073. 8. Guardant lodged the reasons for the action in the UK on 18 August 2025. 9. On 29 August 2025, Guardant lodged an Application for provisional measures before the Paris Local Division against Sophia, for alleged infringement of the patent at issue and EP’533, EP’986 and EP‘073 by offering and supplying the attacked embodiment. 10. At the stage of the Reply to the Objection, Guardant withdrew its request regarding EP’533. 11. In the impugned order, the Paris Local Division “noted” the withdrawal of the request under EP’533 and rejected the Application for provisional measures under EP’073, EP’066 (the patent at issue) and EP’986. Regarding the patent at issue, the Court of First Instance (CFI) held that the patent was likely to be invalid for reasons of added matter. 12. The Paris Local Division ordered Guardant to pay an interim costs award amounting to EUR 400,000. 13. Guardant appealed the order only regarding EP’066 and costs. Sophia lodged a cross-appeal. 14. On 17 February 2026, the Paris Local Division set a deadline of 10 March 2026 for payment of the interim costs award. 15. On 18 February 2026, the Court of Appeal rejected Guardant’s application for suspensive effect with respect to the interim costs award order. **PARTIES’ REQUESTS** 16. In summary, Guardant requests in its appeal that the Court of Appeal set aside the impugned order regarding the patent at issue and the costs and that the Court of Appeal A. order Sophia, in the territories of Austria, Belgium, Bulgaria, Denmark, Estonia, Finland, France, Germany, Italy, Latvia, Lithuania, Luxembourg, Malta, the Netherlands, Portugal, Romania, Slovenia, Sweden, Switzerland, and Spain to cease and desist from I. using and/or offering for use the method according to claims 1 to 3, 6 to 12 and 14 of the patent at issue (direct infringement), and II. supplying and/or offering to supply means, which are suitable and intended for use in a method according to claims 1 to 3, 6 to 12 and 14, specifically i. the software for accessing the “SOPHiA DDMTM platform (II.1i.), ii. the library preparation and hybridization capture kit and components thereof, including but not limited to the “SOPHiA Genetics CUMINTM adapters, “Probes by SOPHiA Genetics” and Instruction for use, in each case i. and ii. above without • In the case of an offer, expressly and clearly indicating that the means may not be used without the consent of Guardant as the proprietor of the patent at issue for detecting variants using the MSK-ACCESS powered with SOPHiA DDMTM according to request A.II., • In the case of supply, imposing on the purchasers, subject to a contractual penalty payment to Guardant of EUR 10,000 for each case of infringement, a written obligation not to use the means for detecting variants using the MSK-ACCESS powered with SOPHiA DDMTM test according to request A.II without prior consent of Guardant as the patent proprietor of the patent at issue, iii. alternative to request A.I, the MSK-ACCESS powered with SOPHiA DDMTM test. B. order Sophia to deliver up, at their own expense, any physical means referred to under A. in stock and/or otherwise held, owned, or in a direct or indirect possession of Sophia in Austria, Belgium, Bulgaria, Germany, Denmark, Estonia, Finland, France, Italy, Latvia, Lithuania, Luxembourg, Malta, the Netherlands, Portugal, Romania, Slovenia, Sweden, Switzerland, and Spain, within one week after service of this order, and to provide Guardant’s counsel with proper evidence of the full and timely compliance with this order within 10 days after the delivery to the bailiff. C. order that for each individual violation of the orders under A. and B., the respective respondent pay to the court a penalty payment of up to EUR 10,000. Each infringing act in relation to MSK-ACCESS powered with SOPHiA DDMTM in respect of request A.I and/or any means in respect of request A.II will be considered as a separate violation. Further, in the case of continuous non-compliance or continuous infringement such as the offering on the internet or non-compliance with the obligation under B., the respective respondent shall pay to the court a penalty payment of up to EUR 100,000 per day. D. order Sophia to bear Guardant’s legal costs in these appeal proceedings in addition to Guardant’s legal costs from the first instance proceedings; E. determine the value of the proceedings to be EUR 1,500,000; F. grant an interim award of costs of EUR 200,000 pursuant to R. 211.1 (d) [RoP]; G. order that the overpaid fees in the amount of EUR 45,100 be reimbursed to Guardant pursuant to R. 371.4 RoP; and H. declare any order to be immediately enforceable. 17. In summary, Sophia requests that the Court of Appeal (i) reject Guardant’s appeal; (ii) order Guardant to bear Sophia’s legal costs in the appeal proceedings, in addition to Sophia’s legal costs from the first instance proceedings; (iii) grant an interim award of costs of EUR 200,000 to Sophia, pursuant to R. 211 (d) RoP and set a deadline for payment of such an award; (iv) In the alternative to requests i to iii, order that the alleged infringement be allowed to continue subject to the provision of a security by Sophia within two weeks, the amount of which is left to the discretion of the Court (but which should not exceed the value in dispute). The security can be provided in the form of a bank guarantee; (v) In the alternative to the requests i to iv, order that the enforcement of the order for provisional measures be dependent on the provision of security by Guardant in the amount of at least EUR 12 million, whereby the security can be provided in the form of a bank guarantee; (vi) With respect to any order for provisional measures, order that enforcement be stayed for a period of three months, to allow hospitals, labs and research institutions to transition to another liquid biopsy product so that patients’ access to critical healthcare is not interrupted. (vii) With respect to all requests, that Guardant be ordered to pay the costs of the proceedings including the costs incurred by Sophia in relation to Guardant’s failed Application for suspensive effect; (viii) With respect to all alternative requests, order that any order for injunctive relief be limited to the Contracting Member States where the patents (respectively) are in force, or in the further alternative, be limited to the Contracting Member States and other EU states where the patents (respectively) are in force. (ix) In the event the Court agrees with Guardant’s lower valuation of the appeal, Sophia also requests reimbursement of the overpaid court fees for the Cross-Appeal, in the amount of EUR 45,100. (x) If (and only if) the Court finds that Sophia’s cross-appeal arguments on urgency and/or necessity are inadmissible as cross-appeal grounds, the Court should direct that Sophia’s submissions on urgency and necessity (as currently pleaded in the cross-appeal) are to be treated as part of Sophia’s Statement of response to appeal. 18. Regarding the cross-appeal, in summary, Guardant requests the dismissal as inadmissible, or alternatively as unfounded. 19. Guardant requests the dismissal of new arguments raised by Sophia on the alleged invalidity of the patent in suit as late filed. 20. Sophia objects para. 24-26 and 28, para. 40a, para. 75, para 87 and 91, para. 88 to 92 and para. 94 in the statement of response to cross-appeal as late filed. **PARTIES’ SUBMISSIONS** 21. In summary, Guardant submits the following: - With regard to urgency Guardant agrees with the CFI’s findings that a three-month period constitutes a reasonable delay to prepare the Application by gathering the necessary evidence. However Guardant emphasises that Guardant did not have sufficient knowledge and also had no possibility to acquire sufficient knowledge to file an infringement action for the UPC in May 2025 or any time before 13 August 2025. - The decision was a surprise because the CFI’s agenda for the oral hearing had not included added matter with respect to the patent at issue. - A method using 25 genes for the four cancers listed in patent claim 1 is precisely disclosed in example 2 of the patent application as filed. - The CFI made a manifest error in awarding interim costs to Sophia without making an order for costs and, more importantly, without Sophia having requested an interim award of costs. - The CFI failed to establish the amount of costs (approximately) incurred. - If the Court is minded to maintain an interim award of costs for the first instance proceedings, Guardant requests that the following points are considered: - An interim award ceiling for the first instance proceedings should in fact be half of three quarters of the cost ceiling of EUR 600,000, so half of EUR 450,000, which is EUR 225,000. At most, if the cost ceiling is considered to be EUR 600,000 then the interim award ceiling should be EUR 300,000. - Sophia’s failure to respond appropriately to Guardant’s correspondence regarding the patent that was withdrawn, caused unnecessary costs that should be borne by Sophia (Art. 69(3) UPCA). - Sophia’s costs of producing the invalid submission of 9 December 2025 were unnecessary, and should be borne by Sophia. - For all these reasons, even if interim costs are awarded, Guardant requests that the Court of Appeal set the ceiling for any interim award of costs at EUR 225,000 for Sophia. 22. Essentially, Sophia defends the impugned decision and submits the following: - The assessment of urgency of the CFI is incorrect. - With regard to necessity Sophia argues that Guardant did not provide any evidence in support of the allegation that allowing the attacked embodiments to stay in the market would cause significant damages to Guardant. In addition there is a fundamental imbalance in the positions of Sophia and Guardant. - Not filing a merits action at the UPC also suggests a lack of necessity on Guardant’s part. - The lack of urgency and necessity on Guardant’s part can also be seen through their resistance to progressing the parallel case in the UK. To date, it is Sophia who has had to drive the listing of the case management conference and trial in the UK. - The UPC lacks international jurisdiction regarding Spain and Switzerland. **REASONS** **A. International Jurisdiction** 23. Sophia contests the jurisdiction of the UPC with respect to acts of alleged infringement in Spain and Switzerland. 24. As Sophia Genetics SAS, Sophia Genetics SRL and Sophia Genetics GmbH are domiciled in Contracting Member States of the UPC, namely France, Italy and Germany, the jurisdiction concerning this application follows from Art. 4 in conjunction with Art. 71b(1) Regulation 1215/12 also insofar it relates to Spain and Switzerland. Jurisdiction pursuant to these provisions is not limited to acts of the defendant within the UPC Territory. The connection with the UPC Territory is established by the domicile of the defendant, and not the location where the acts of the defendant occurred (CoA, referral order 6 March 2026, UPC CoA 789/2025, UPC CoA 813/2025, Dyson v Honeywell, para. 12). 25. Regarding Sophia Genetics SA domiciled in Switzerland, Guardant claims that this company is jointly responsible with Sophia Genetics SAS, France, for Sophia’s website. Guardant argues that for Sophia Genetics SA, jurisdiction follows from Art. 8(1) Regulation 1215/12, with Sophia Genetics SAS, France, being the anchor defendant. 26. The Court of Appeal agrees. In the case-law of the CJEU, it has been clarified that Art. 8(1) Regulation 1215/2012 must be interpreted as meaning that a situation where two or more companies are each separately alleged to have committed an infringement on the same national part of a European patent by carrying out reserved actions with regard to the same product is capable of giving rise to “irreconcilable judgments” resulting from separate proceedings as referred to in that provision (see, in relation to the similar provision of Art. 6(1) of Regulation (EC) No 44/2001, CJEU, 12 July 2012, Solvay v Honeywell, C-616/10, ECLI:EU:C:2012:445; CoA, Dyson v Dreame, para. 16). In this case Guardant even alleges that Sophia Genetics SA (Switzerland) and Sophia Genetic SAS (France) are jointly responsible for the website and for distributing the software in Spain and Switzerland. **B. Subject matter of the patent at issue** **a) The patent and its technical background** 27. The patent at issue relates to a method for detecting the presence or absence of cancer. 28. According to the patent description, several screening tests to detect cancer are available (para. 2 line 9), in particular genetic tests. Genetic tests detect certain gene deleterious mutations linked to some types of cancer. Genetic testing is particularly useful for a number of diagnostic methods (para. 2 lines 14-15). 29. According to the description, with improvements in sequencing and techniques to manipulate nucleic acids, there is a need in the art for improved methods and systems for using cell free DNA (“cfDNA”) to detect and monitor disease. In particular, there is a need to balance efficient use of sequenced base pairs with high accuracy and sensitivity variant detection (para. 64). 30. Next generation sequencing allows detection and measurement of rare mutations. Detection of mutations relative to germline sequence in a fraction of cell-free DNA can indicate the presence of ctDNA (circulating tumor DNA), thus indicating the presence of a tumor. Sequencing cell free DNA may allow detection of mutations in cancer related genes (para. 82 lines 3-7). **b) Objective of the invention** 31. Against this background, the problem underlying the invention is to provide an improved method for using cfDNA to detect and monitor cancer. **c) Feature breakdown** 32. This objective is achieved by a method with the following features (using the feature breakdown established by the Paris Local Division). 1.1 A method for detecting the presence or absence of colorectal cancer, ovarian cancer, lung cancer or pancreatic cancer in a subject comprising: 1.2 sequencing circulating cfDNA from the subject 1.3 at a depth of at least 50,000 reads per base to detect one or more genetic variants associated with cancer, 1.4 wherein the sequencing is performed on an enriched set of amplified cfDNA molecules 1.5 which comprises a panel of genomic regions, wherein the genomic regions in the panel comprise one or more loci from each of the genes AKT1, ALK, APC, ATM, BRAF, CTNNB1, EGFR, ERBB2, ESR1, FGFR2, GATA3, GNAS, IDH1, IDH2, KIT, KRAS, MET, NRAS, PDGFRA, PIK3CA, PTEN, RB1, SMAD4, STK11 and TP53, and 1.6 further comprising amplifying the cfDNA prior to sequencing, and 1.7 determining a consensus sequence from sequence reads obtained from the sequencing to reduce errors from amplification or sequencing. **d) Person skilled in the art** 33. The person skilled in the art is a team consisting of a molecular biologist, familiar with Next-generation sequencing technology (“NGS”) and genetic testing, and at least one oncologist. The NGS is used to determine the order of nucleotides in entire genomes or targeted regions of DNA or RNA, wherein a big number of fragments are sequenced at the same time. **e) Interpretation of patent claim 1** 34. The Court bases its decision on the following understanding of the features: **Feature 1.1** 35. Feature 1.1 requires “a method for detecting the presence or absence of colorectal cancer, lung cancer or pancreatic cancer in a subject”. 36. As this is a method claim, the scope of the patent is limited to the use for that specific purpose. **Identification of the cancer type necessary?** 37. Sophia argues without success that it is not enough for the claimed method to provide a result which means that the subject may have one or more cancers listed in the claim, without identifying which of those cancers are present or absent. 38. No such restriction can be inferred from the wording of this feature. According to the wording, the subject of the detection is the presence or absence of the specified types of cancer, rather than the cancer type itself. 39. This also finds support in the description. There it is stated that methods according to the claimed invention are used to detect cancer in a subject. Cell free DNA can be sequenced in subjects not known to have cancer or suspected of having cancer to diagnose the presence of [sic] absence of a cancer (para. 96 lines 26-28, emphasis added by the Court). That the patent description distinguishes between diagnosing a cancer and determining its specific type, thus, information about the cancer, is clear from the following sentences in the same paragraph: “Sequencing cell free DNA provides a non-invasive method for early detection of cancer or for ‘biopsy’ of a known cancer. Cell free DNA can be sequenced in subjects diagnosed with cancer to provide information about the cancer. Cell free DNA can be sequenced in subjects before and after treatment for cancer to determine the efficacy of the treatment” (para. 96 lines 28-31, emphasis added by the Court). 40. Against this background, contrary to Sophia’s view, the fact that paragraphs 111, 112 and 114 of the description of the patent at issue refer to panels of genes for each of the four cancers listed in claim 1 does not support the interpretation that feature 1.1 requires the identification of the cancer type. 41. Paragraph 183 lines 36-41, to which Sophia also refers, does not either. It reads as follows: “In some embodiments, other regions are targeted that are stable and/or not differently regulated in tumor cells. Within these regions, differences in coverage, cleavage sites, fragment length, sequence content, sequence content at fragment endpoints, or sequence content of the nearby genomic context can be used to infer the presence or absence of a certain classification of cancer cells (e.g., EGFR mutant, KRAS mutant, ERBb2 amplified, or PD-1 expression cancers), or type of cancer (e.g., lung adenocarcinoma, breast or colorectal cancer)”. The detection of the presence or absence of cancer and the identification of the type of cancer are listed here as alternatives. Furthermore, this paragraph only concerns embodiments. Embodiments or examples cannot generally be understood as limiting the scope of the claim (see CoA, 3 March 2025, UPC CoA 523/2024, Sumi Agro v Syngenta, para. 52). 42. Sophia’s opinion is not forced by the fact that the specification (para. 100) mentions over 100 different types of cancer. This paragraph needs to be read in context with the previous paragraphs. Paragraphs 97-99 explain for which type of subjects the claimed method may be used, such as a subject suspected of having cancer (para. 97) or being at risk of having cancer (para. 99). Para. 100 explains it may be used in a situation where the subject may already have been diagnosed with a cancer, the 100+ different types of cancer being summed up as examples of the type of cancer such subject may have been diagnosed with. **Detection of all four cancer types or differentiation between them necessary?** 43. Feature 1.1. mentions the four cancer types as alternatives (“colorectal cancer, ovarian cancer, lung cancer or pancreatic cancer”, emphasis by the Court). Hence, for the realization of feature 1.1, it is sufficient that the claimed method is used to detect the presence or absence of one of the listed cancer types. It is not necessary that the method is used to differentiate between the individual cancer types (“differential diagnosis”). 44. Even if it were to be considered, the prosecution history does not suggest otherwise. Guardant argued before the EPO: “This single panel therefore provides an effective test across multiple cancer types, thus negating the need for a different test to identify each of these cancers (SG 92, page 3)”. This only means that the panel in feature 1.5 can, in principle, detect the presence of cancer types named in feature 1. But it does not necessarily mean that claim 1 requires that the test is used to identify each of these individually. **Separation from all other kind of cancer and specific sensitivity necessary?** 45. According to Sophia, feature 1.1 teaches that the claimed test can distinguish between patients with the listed four specific cancer types and all patients without these four specific cancer types. This is not correct. As stated above, it is sufficient if the use of the claimed method is aimed to detect one of the listed cancer types in a subject. Already for this reason there is no suggestion that the diagnosis should distinguish between groups of patients, as Sophia suggests. 46. The person skilled in the art understands that, due to feature 1.5, the test provides increased sensitivity for the four specific cancer types listed in feature 1.1 and therefore focuses on these cancer types. The purpose of the test is not to distinguish from other cancer types. From para. 245, the person skilled in the art understands that the invention provides “assay improvements for increased sensitivity” (para. 245 line 46). The paragraph explains that “the CRC-specific panel was expanded into a new panel designed to achieve high sensitivity for colorectal, ovarian, lung, and pancreatic cancers. Sites where mutations were prevalent in the GH database were also included. This 25-gene panel reported SNVs in 25 genes, indels in 7 genes, and fusions in 1 gene. FIG. 15 shows genes selected for detection of major cancer types with >90% theoretical sensitivity” (para. 245 lines 46-49). FIG. 15 shows all the genes mentioned in feature 1.5. 47. Although feature 1.5 is intended to achieve particularly high sensitivity, this does not mean, as Sophia believes, that a specific level of sensitivity must be achieved to fulfil feature 1. **Feature 1.2** 48. The step according to feature 1.2 is sequencing the circulating cfDNA from the subject. 49. DNA molecules are composed of a series of four different nucleotides (known as “A”, “C”, “G” and “T”) and a DNA molecule can be represented by a sequence of these four letters. DNA sequencing is the process by which a DNA molecule is analysed to reveal the series of its nucleotides. **Feature 1.3** 50. Feature 1.3 stipulates that the sequencing must be performed ‘at a depth of at least 50,000 reads per base to detect one or more genetic variants associated with cancer’. 51. Modern sequencing machines (or simply ‘sequencers’) can determine the sequence of many millions of different molecules in parallel. 52. Before DNA molecules can be sequenced in a NGS machine they must first be converted into a ‘sequencing library’. This involves inter alia adding ‘sequencing adapters’ to the ends of the DNA molecules. 53. The sequence library is then processed in the NGS machine, and the basic output from this process is a set of ‘sequence reads’. These are short strings of letters (A, C, G, T) which represent the DNA sequence for a single sequencing reaction in the machine. 54. The human genome is 3 billion nucleotides long, and a single cell (except for gametes) includes two copies of the genome, one inherited from each parent i.e. 6 billion nucleotides. If sequence reads are 200 nucleotides, then 30 million sequence reads are needed to cover a single cell. 55. The number of times that a particular nucleotide is seen is referred to as its ‘sequencing depth’. 56. It follows that feature 1.3 requires that a particular nucleotide must be sequenced 50,000 times and has to be seen in 50,000 sequence reads. Sequencing of every region of the panel at a depth of 50,000 reads per base (rpb)? 57. Guardant argues that the claim does not state that every region in the 25 genes (feature 1.5) must be sequenced at a depth of at least 50,000 rpb. Sophia objects this successfully as incorrect reading of the claim. Feature 1.3 requires that all loci of the 25 genes included in the panel must be sequenced with a minimum of 50,000 rpb. 58. Claim features must always be interpreted in light of the claim as a whole (CoA, 13 May 2024, UPC 1/2024 para. 29, VusionGroup SA v Hanshow; 27 May 2026, UPC CoA 622/2025, UPC CoA 623/2025, Hefei v Grundfos, para. 58). Therefore feature 1.3 has to be read together with feature 1.4. Feature 1.4 states that subject of the sequencing is the enriched set of amplified cfDNA molecules. That means that not only a portion of the panel has to be sequenced. 59. Neither the claim nor the description indicates that it is sufficient for only parts/regions of the panel to be sequenced at a minimum of 50,000 rpb. To the contrary, many parts of the description indicate that the chosen panel is sequenced at a certain depth. For example, it is described that ”the panel may be selected to limit a region for sequencing to a fixed number of base pairs. The panel may be selected to sequence a desired amount of DNA. The panel may be further selected to achieve a desired sequence read depth. The panel may be selected to achieve a desired sequence read depth or sequence read coverage for an amount of sequenced base pairs” (para. 107 lines 42-45, see also para. 153). 60. Nothing else can be inferred from ‘comprise’ in feature 1.5. At least one locus of each of the genes must be included in the panel, but whatever is included needs to be sequences with the required depth. **Features 1.4 and 1.6** 61. According to feature 1.4 the sequencing must be performed on an enriched set of amplified cfDNA molecules. 62. Amplified (see also feature 1.6: “amplifying”) means that the amount of DNA is enhanced, for example by polymerase chain reaction (PCR), whereby portions of DNA can be repeatedly copied in an exponential manner, vastly increasing the amount which is available for analysis. **Feature 1.5** 63. According to feature 1.5 the enriched set of amplified cfDNA molecules comprises a panel of genomic regions, wherein the genomic regions in the panel comprise one or more loci from each of the genes listed in feature 1.5. 64. This does not require that only loci of genes listed in feature 1.5 are in the genomic region. This follows from the “comprising-of” language. **Feature 1.7** 65. According to feature 1.7 the consensus sequence must be determined from the sequence reads obtained from the sequencing to reduce errors from amplification and sequencing. 66. Where a DNA molecule has been sequenced multiple times, the various sequence reads can be compared to generate “calls” that represent the best prediction (or consensus) for the true identity of the nucleotide at each position in that molecule. Differences between sequence reads for the same molecule (e.g. due to experimental noise) are thus removed. 67. A consensus sequence is therefore the “most likely” nucleotide position for an original molecule. This is undisputed between the parties and well known in the art. If for example 95 out of 100 reads at a particular position give nucleotide A, the consensus at that position is A. 68. Feature 1.7 requires that such a consensus sequence be determined. **Collapsing** 69. The parties have different opinions on the question whether feature 1.7 requires collapsing. Both parties argue that the other party raised this point for the first time during the appeal proceedings and that the argument should therefore not be admitted. **New arguments** 70. There is no reason to disregard these arguments. Pursuant to R. 222.2 RoP, the Court of Appeal may disregard “requests, facts and evidence” which were not submitted by a party during the proceedings before the Court of First Instance. This wording makes it clear that the rule does not apply to legal arguments. R. 222.2 RoP therefore does not prevent a party from submitting a new legal argument on appeal, provided that the argument is based on facts and evidence submitted to the Court of First Instance (CoA, 21 November 2024, UPC CoA 511/2024, Meril v Edwards, para. 12). The interpretation of a patent claim is a matter of law (CoA, 30 April 2025, UPC CoA 768/2024, Insulet v EOFlow, para. 37; 19 June 2025, UPC CoA 402/2024, para. 18). 71. Sophia did not bring new facts and evidence with regard to the interpretation of feature 1.7. Sophia argued in the Objection that the term “consensus sequence” can be construed in the same way in the patent at issue as in EP ́073 and referred to para. 196 to 200 of the Objection. There Sophia explained that the term “consensus sequence” is a term of art. 72. Sophia argues unsuccessfully that Guardant for the first time argues that the consensus sequence in the patent at issue is different to that in EP ́073. That is not true. Guardant in the Reply to the objection argued as follows: “Finally, with respect to feature 1.7, the defendants refer to [211] of the 066 patent (...), which is similar to [0078] in the 073 patent, so the teaching of this paragraph concerning the use of probabilities to reduce noise has already been addressed above. However, the 066 patent goes further than 073 (...) and specifies that the consensus sequence can be performed on a ‘base-by-base basisʼ (...). Moreover, [0168] expressly describes using a probabilistic model” (emphasis by the Court). Collapsing necessary or probablistic functions sufficient? 73. Collapsing means that the cfDNA molecules are grouped together. 74. Sophia argues unsuccessfully that the act of collapsing the family of consensus reads coming from the same original starting molecule is inherent to producing a consensus sequence for that original starting molecule. 75. Claim 1 leaves it open how the consensus sequence is determined. Unlike claim 1 of EP ́073, feature 1.7 does not specify that the determination of the consensus sequence involves a step of collapsing. Hence, each method to determine the “most likely” nucleotide position for an original molecule is possible. 76. The person skilled in the art infers from paragraph 211 that “grouping sequence reads into families derived from individual molecules” – and thus “collapsing” – is only one of two ways to reduce noise and/or distortion from a single individual molecule or from an ensemble of molecules. This is the purpose of feature 1.7 (“to reduce errors from amplification or sequencing”). 77. Para. 211 reads as follows: “One way to reduce noise and/or distortion from a single individual molecule or from an ensemble of molecules is to group sequence reads into families derived from original individual molecules to reduce noise and/or distortion from a single individual molecule or from an ensemble of molecules. (...) Additionally, high yield generation of consensus sequences for tagged parent polynucleotides by high-rate sampling of progeny polynucleotides amplified from the tagged parent polynucleotides, and collapsing of generated sequence reads into consensus sequences representing sequences of parent tagged polynucleotides can reduce noise introduced by amplification bias and/or sequencing errors, and can increase sensitivity of detection. Collapsing sequence reads into a consensus sequence is one way to reduce noise in the received message from one molecule. Using probabilistic functions that convert received frequencies into likelihood or posterior estimates of each of the possible true nucleotides using defined estimates of amplification and sequencing error profiles is another way to reduce noise and/or distortion” (emphasis by the Court). 78. That the consensus sequence is only mentioned in para. 211 regarding the collapsing of generated sequence reads, does not mean that feature 1.7 is limited to this method. The wording of the claim is broader and only requires ‘determining a consensus sequence from reads obtained from sequencing (...)’. The person skilled in the art recognises that this refers to any method suitable for determining a consensus sequence, i.e. the “most likely” nucleotide position for an original molecule. 79. The fact that the consensus sequence can be determined by using probabilistic functions follows also from para. 23, lines 50-52, where it is stated that “in some embodiments detection of consensus sequence is based on assessing probabilities of each of the potential nucleotides based on the observed sequencing output (...)” (emphasis by the Court). 80. Sophia refers without success to para. 168, which reads as follows: “To reduce error, sequence reads may be collapsed to generate a set of consensus sequences. To generate consensus sequences, sequence reads may be collapsed based on sequence information in the non-unique tag and endogenous sequence information, including sequence information at a beginning region of a sequence read, sequence information at an end region of a sequence read, and a length of as sequence read. (...) A consensus sequence can be determined on a molecule-by-molecule basis (wherein a consensus sequence is determined over a stretch of bases) or a base-by-base basis (wherein a consensus nucleotide is determined for a base at a given position).” 81. Sophia argues that it follows from para. 138, in particular “to generate consensus sequences, sequence reads may be collapsed to generate a set of consensus sequences”, that collapsing is inherent to the creation of consensus nucleotide. Given that the patent description describes the use of probabilistic functions as an alternative means of determining the consensus sequence, a limitation to collapsing is not justified. Consensus sequence reached for each cfDNA molecule (“molecule-by-molecule basis”) or on a “base-by-base basis”? 82. Claim 1 leaves it open whether the consensus sequence is determined for each cfDNA molecule, for the whole molecule, for a certain region of a molecule, or even only for single loci of a molecule. Paragraphs 23 and 168 provide a very broad and general description of how a consensus sequence can be determined, indicating that determining the consensus sequence can be performed on a molecule-by-molecule basis (wherein a consensus sequence is determined over a stretch of bases) or a base-by-base basis (wherein a consensus nucleotide is determined for a base at a given position). **C. Added matter** 83. Guardant objects successfully the finding of the Paris Local Division that the patent at issue is more likely than not invalid on the grounds of added matter. **Principles** 84. Under Article 138(1)(c) EPC a European patent may be revoked if its subject-matter extends beyond the content of the application as filed or, if it was granted on a divisional application, extends beyond the content of the earlier application as filed. 85. In order to ascertain whether there is added matter contrary to Art. 123(2) EPC, the Court must first ascertain what the skilled person would derive directly and unambiguously using his common general knowledge and seen objectively and relative to the date of filing, from the whole of the application as filed, whereby implicitly disclosed subject-matter, i.e. matter that is a clear and unambiguous consequence of what is explicitly mentioned, shall also be considered as part of its content (CoA, 14 February 2025, UPC CoA 382/2024, Abbott v Sibio, para. 52; 25 November 2025, UPC CoA 528/2024, Amgen v Sanofi, para. 54). 86. On the basis of these principles, all features of claim 1 of the patent at issue are disclosed in the application as filed (WO 2017/1811146 A1, Exhibit SG94), contrary to the view of the Paris Local Division. **Disclosure of features 1.2 to 1.5** 87. Features 1.2 to 1.5 are directly and unambiguously disclosed in claim 49 in connection with claims 44 and 37 and the description. Claim 49 depends only on claim 44, which on its turn depends only on independent claim 37. 88. Claim 37 of the application concerns a method for detecting cancer in a subject comprising: sequencing circulating cell-free DNA (cfDNA) from the subject (feature 1.2) at a depth of at least 50,000 reads per base to detect one or more genetic variants associated with cancer (feature 1.3). 89. Subject of claim 44 of the application is a method according to claim 37, wherein the sequencing is performed on an enriched set of cfDNA molecules (feature 1.4). According to the description, methods disclosed therein can comprise amplifying of polynucleotides (SG94, para. 235). 90. Subject of claim 49 is the method of claim 44, wherein the enriched set of cfDNA molecules comprises one or more genes selected from the group consisting of the genes listed in feature 1.5 of claim 1 of the patent at issue (emphasis by the Court). That claim 49 in contrast to feature 1.5 only mentions “genes” and not “loci” of these genes does not matter, because the skilled person understands gene to mean either the full gene, or one or more loci of such gene. That is because the application discloses that for improving the likelihood of detecting tumor indicating mutations, the region of DNA sequenced may comprise not only a panel of genes but also genomic regions (SG94, para. 155). It is expressly stated in the description of the application that in some cases, the one or more regions in the panel can comprise “one or more loci from one or a plurality of genes, including one or more of AKT1, ALK, APC, ATM, BRAF, CTNNB1, EGFR, ERBB2, ESR1, FGFR2, GATA3, GNAS, IDH1, IDH2, KIT, KRAS, MET, NRAS, PDGFRA, PIK3CA, PTEN, RB1, SMAD4, STK11, and TP53” (SG94, para. 159), i.e. the 25 genes of feature 1.5. 91. The person skilled in the art concludes from claims 37 and 49 a sequencing depth of at least 50,000 rpb for such a panel (feature 1.3). **Disclosure of feature 1.1** **Detecting the presence or absence of cancer** 92. Claim 37 of the application reads as follows: “A method for detecting cancer in a subject comprising: sequencing circulating cell-free DNA (cfDNA) from the subject of at least 50,000 reads per base to detect one or more genetic variants associated with cancer”. 93. Contrary to Sophia’s opinion, the insertion of “the presence or absence” in feature 1.1 of claim 1 of the patent at issue does not lead to added matter. The person skilled in the art derives from para. 145 of the Application, which is identical to para. 96 of the patent at issue, that the detection of cancer has the same meaning as the diagnosis of the presence or absence of cancer. This is because it reads as follows: “Methods herein can be used to detect cancer in a subject. Cell free DNA can be sequenced in subjects not known to have cancer, or suspected of having cancer to diagnose the presence of [sic.] absence of a cancer. Sequencing cell free DNA provides an noninvasive method for early detection of cancer or for ‘biopsy’ of a known cancer” (emphasis by the Court). **Disclosure of feature 1.7** 94. At the oral hearing before the Court of Appeal, Sophia argued that feature 1.7 is not disclosed in the application as filed. This is not true. According to the description of the application, “In some embodiments the method further comprises amplifying the cfDNA prior to sequencing, and determining a consensus sequence from sequence reads obtained from the sequencing to reduce errors from amplification or sequencing” (SG94, para. 37). Furthermore, para. 218 of the application states (like para. 168 lines 39-48 in the patent at issue): “To reduce error, sequence reads may be collapsed to generate a set of consensus sequences. To generate consensus sequences, sequence reads may be collapsed based on sequence information in the non-unique tag and endogenous sequence information, including sequence information at a beginning region of a sequence read, sequence information at an end region of a sequence read, and a length of as sequence read. (...) In some embodiments, a probabilistic model is constructed to model amplification and sequencing error profiles and used to estimate probabilities of true nucleotide in each position of the molecule” (SG94, para. 218 lines 5-16). 95. Thus, for the skilled person, using amplification goes together with determining consensus sequence in order to reduce errors. He sees this confirmed by claim 56, which depends on claim 37, adding amplification as well as determining consensus sequence to the method of claim 37. 96. It is also disclosed that consensus sequencing is not only achieved by way of collapsing sequence reads. The application has the following content: “Collapsing sequence reads into a consensus sequence is one way to reduce noise in the received message from one molecule. Using probabilistic functions that convert received frequencies into likelihood or posterior estimates of each of the possible true nucleotides using defined estimates of amplification and sequencing error profiles is another way to reduce noise and/or distortion” (emphasis added, SG94, para. 264 midway, similar to para. 211 of the patent at issue). **Cancer types listed in feature 1.1** 97. It is true that claim 37 does not specify the type of cancer. But example 2 does. **Reliance on Example 2?** 98. During the CFI proceedings, Guardant relied on example 2 of the application for the first time at the oral hearing to support its defence against the added matter argument. There is no reason why example 2 in the application should be disregarded. Whether there is added matter is a question of law. As the test is whether the relevant claims are disclosed in the application as a whole, the Court is permitted to consider the entire document (Amgen v Sanofi, para. 61). **Description of Example 2 in the application as filed** 99. Essentially, example 2 is described as follows in the application: “This example demonstrates a study with Colorectal Cancer (CRC) patients and the subsequent assay improvements driven by the results. In this example, earlier stage cancers, with which less tumor DNA was shed into circulation (FIG. 11), and was detected by the analysis of cell-free circulating tumor DNA (ctDNA) with next-generation sequencing (NGS) (SG94, para. 290). As a first iteration of this process, a 25 kb capture panel was developed based on the landscape of genomic alterations in ctDNA of over 10,000 advanced cancer patients (GH database). The panel content was selected to achieve high clinical sensitivities for colorectal (96%), ovarian (95%), lung (87-93%) and pancreatic (88%) (SG94, para. 291, emphasis added by the Court). Panel: The 25-gene panel reported SNVs in 25 genes, indels in 7 genes, and fusions in 1 gene. FIG. 12 shows oncoprints of four major cancer types: colorectal adenocarcinoma, pancreatic adenocarcinoma, lung adenocarcinoma, and ovarian serous cystadenocarcinoma corresponding to a subset of genes on the 25-gene panel“ (SG94, para. 292, emphasis added by the Court). Assay improvements for increased sensitivity: The CRC-specific panel was expanded into a new panel designed to achieve high clinical sensitivity for colorectal, ovarian, lung, and pancreatic cancers. Sites where mutations were prevalent in the GH database were also included. This 25-gene panel reported SNVs in 25 genes, indels in 7 genes, and fusions in 1 gene. FIG. 15 shows genes selected for detection of major cancer types with ˃ 90% theoretical sensitivity” (SG94, para. 298, emphasis added by the Court). 100. The 25 genes of claim 49 are the ones selected for the panel of example 2. Thus the person skilled in the art sees that example 2 shows an example for a method according to claim 49. The skilled person thus links example 2 also to the method of claim 37, on which claim 49 depends. Example 2 clearly links the panel with the selected 25 genes to the detection of at least one of the four cancer types of feature 1.1. 101. Example 2 discloses feature 1.1 directly and unambiguously. 102. As explained in the section on claim interpretation (para. 37-42) feature 1.1. does not require that the type of the listed cancers be determined. 103. Sophia argues without success that there is no information in example 2 on whether the test could be successfully used to differentiate colorectal cancer from any of the 100+ cancers listed elsewhere in the specification that are not included in feature 1.1. As explained above (para. 43) the test is only required to detect one of the four specific cancer types listed in feature 1.1. That is disclosed in the application, in particular in para. 291 where the method used in example 2 is explained, which reads: ”(...) The panel content was selected to achieve high clinical sensitivities for colorectal (96%), ovarian (95%), lung (87-93%) and pancreatic (88%) (...)” (SG94, para. 291, emphasis added by the Court). 104. It is true that the clinical data in example 2 is only directed to colorectal cancer and para. 291 only discusses the theoretical capability of the test. No experiments are performed to confirm that the detection of the presence or absence of ovarian, lung or pancreatic cancer is possible with the panel. Contrary to Sophia’s opinion, this is not necessary. It is sufficient that it is disclosed that the panel content was selected (para. 291)/expanded (para. 298) to achieve high clinical sensitivities for colorectal (96%), ovarian (95%), lung (87-93%) and pancreatic (88%) (SG94, para. 291 and 298, emphasis added by the Court). **Inherent contradiction of example 2** 105. Sophia claims that example 2 has an inherent contradiction. Sophia argues as follows: Para. 290 says that as a first iteration of this process (emphasis added) a 25kb panel was developed to achieve high clinical sensitivities for colorectal (96%), ovarian (95%), lung (87-93%) and pancreatic (88%). Para. 298 then confusingly says “The CRC [colorectal]-specific panel was expanded into a new panel designed to achieve high clinical sensitivity for colorectal, ovarian, lung and pancreatic cancers”. Sophia sees a clear ambiguity, since the panel as originally described in para 291-292 appears to target the same cancers and comprises the same genes as the panel described in para. 298. But the panel in para. 298 is described as “expanded”, “new” panel compared to an earlier “CRC-specific panel”. 106. The ambiguity claimed by Sophia does not exist. Para. 291 clearly states that a panel was developed based on the landscape of genomic alterations in cfDNA of over 10,000 advanced cancer patients. The panel content was selected to achieve high clinical sensitivity for the four cancer types listed in feature 1.1. According to para. 293 “this panel was applied to a study cohort of 21 CRC patients.” As para. 298 then explains that the “CRC-specific panel was expanded into a new panel designed to achieve high clinical sensitivity for colorectal, ovarian, lung and pancreatic cancers”. The person skilled in the art understands that this panel is the same panel as described in para. 291 and 293. “Expanded” means that the full panel is used as opposed to the earlier CRC-test which focused on only part of the panel, as the oncoprints are for a subset of genes as explained in para. 293. The description of the panel in para. 292 and 298 is identical (25-gene panel, indels in 7 genes, and fusions in 1 gene). This panel description corresponds exactly to the panel depicted in Figure 15. If there were a difference between the two panels, a clarification would have been provided. This also means that, contrary to Sophia’s opinion, both FIG. 13 and FIG. 15 depict the panel designed to achieve high clinical sensitivity for colorectal, ovarian, lung and pancreatic cancers. **Disclosure of feature 1.6** 107. FIG. 13, which depicts example 2, shows that the gene panel was amplified before sequencing. Contrary to Sophia’s opinion, para. 293 which references to FIG. 13 does not describe a CRC (colorectal cancer)-specific panel (see above para. 106). **D. Urgency** 108. Sophia successfully objects against the Paris Local Division’s finding that Guardant sought provisional measures within a reasonable delay. 109. When weighing the interests, the Court shall have regard to any unreasonable delay in seeking provisional measures in accordance with R. 211.4 RoP. If the patent proprietor's behaviour shows that the enforcement of his rights is not urgent, provisional legal protection is not required. 110. Delay within the meaning of R. 211.4 RoP shall be calculated from the day on which the applicant became aware, or should have become aware, of the infringement that would enable him, in accordance with R. 206.2 RoP, to file an application for provisional measures with a reasonable prospect of success. Thus, the decisive point in time is when the applicant has, or should have had, after exercising due diligence, the necessary facts and evidence within the meaning of R. 206.2(d) RoP. Whether there has been an unreasonable delay within the meaning of R. 211.4 RoP depends on the circumstances of the individual case (CoA, 25 September 2024, UPC CoA 182/2024, Mammut vs Ortovox). 111. It is the applicant's burden to demonstrate to the Court that it has acted without unreasonable delay (see Court of Appeal, 26 February 2024, UPC CoA 355/2023, Nanostring v 10X Genomics, Headnote 3). 112. The facts submitted and the evidence presented by Guardant are insufficient for the Court of Appeal to conclude that Guardant acted without undue delay. **Relevant starting date** 113. The Court of Appeal is not convinced that Guardant acted with the required urgency when filing the Application for provisional measures. That there is an unreasonable delay does not simply follow from the fact, that in the UK, Guardant started litigation on 14 July 2025, asserting the UK designations of EP‘533, EP‘986 and the patent at issue as well as a national patent in the same family as EP’073. As Guardant lodged the reasons for the action in the UK not before 18 August 2025, it cannot be assumed that, on 14 July 2025, Guardant was already in a position to apply for preliminary measures with regard to the patent at issue before the UPC. The Court is for other reasons convinced that Guardant should have been aware of the infringing acts within the UPC territory, Switzerland and Spain on 15 May 2025. This follows from the following: **Knowledge of, or opportunity to gain knowledge of the alleged realisation of the patent claims** 114. In this context it is relevant when Guardant became aware of an infringement of the patent claims in question; it is not relevant when Guardant acquired the necessary knowledge regarding the other patents asserted at first instance. A patent holder is not obliged to assert all patents in one application for provisional measures. If a patent holder has the necessary information for filing an application for provisional measures regarding some, but not all patents, delaying the filing of the Application for provisional measures until it has the necessary information regarding all infringed patents may be regarded as an unreasonable delay. 115. Neither the UPCA nor the RoP provides for the need to lodge infringement claims or an Application for provisional measures relating to all infringed patents in a single action or Application. Contrary to Guardant’s opinion, it is compatible with procedural efficiency (point 4 of the Preamble of the RoP) and a frontloaded system to file multiple separate Applications involving the same attacked embodiments, the same parties and legal grounds, particularly with regard to necessity and urgency, weeks apart before the same division. This is evident from the fact that asserting multiple patents in a single Application carries the risk that a prompt decision cannot be expected, while an application for provisional measures is characterised by the fact that a decision is required urgently. 116. Guardant had knowledge or should have had knowledge of the alleged realisation of the claims of the patent at issue by the attacked embodiments by 1 May 2025. This is demonstrated by the following. 117. With regard to all four patents asserted before the CFI, Guardant primarily relied on the User Manual (Exhibit GH21) and the Instructions for Use (Exhibit GH22) to demonstrate infringement by the attacked embodiments. With regard to the alleged infringement of EP’986 Guardant mainly relied on GH21 as well as on Exhibit GH39. **User Manual (Exhibit GH21)** 118. Guardant was aware of the User Manual (Exhibit GH21) prior to 27 May 2025, the date of the UK warning letter. This is evidenced by Guardant's reference to the Manual in the warning letter dated 27 May 2025. Sophia successfully argues that Guardant had to perform the necessary analysis to make the assertions set out in the warning letter. Guardant was therefore aware or should have been aware of all the details relevant to the infringement of the patent in dispute well before 27 May 2025. 119. Contrary to Guardant's view expressed during the Court of Appeal hearing, the fact that the warning letter did not concern the patent at issue does not justify the conclusion that Guardant was unaware of, or had no reason to be aware of, the infringement of the patent at issue prior to sending the warning letter. If a patent holder is aware, on the basis of a document, that one or more of its patents has been infringed, it must not turn a blind eye to the fact that the document also indicates the infringement of one or more of its other patents. 120. That Guardant knew all the relevant details of the attacked embodiments in May 2025 is also evident from the written statement by its Intellectual Property Vice-President (Exhibit GH40), who declares inter alia: “(...) no senior executive of Guardant, nor any employee with the responsibility for evaluating potential intellectual property or patent infringement, was aware of: The existence of Sophia's commercial activities in relation to the ˈMSK-ACCESS powered with SOPHiA DDMˈ assay anywhere in Europe or The technical details of the ˈMSK-ACCESS powered with SOPHiA DDMˈ assay before May 2025 (emphasis by the Court)”. 121. As Guardant bears the burden of proof in this regard and has not provided any precise details of the date on which the relevant persons became aware of the relevant technical details, especially when Guardant got access to GH21, it must be assumed that the technical details relevant for the patent at issue were known by 1 May 2025. **Instructions of Use (Exhibit GH22)** 122. Nothing else can be inferred from Guardant's allegations that only on 11 July 2025, Guardant obtained the Instructions for Use (Exhibit GH22). Guardant has not credibly demonstrated that it did not receive or could have received this document before 11 July 2025. The Vice-President for Intellectual Property's testimony (Exhibit GH40) makes no reference to Annex GH22 or the date of 11 July 2025. 123. Furthermore, the only decisive factor is that Guardant had obtained sufficient knowledge of the technical details to justify the infringement assessment from GH21. 124. Regarding feature 1.2, in the Application, Guardant not only relied on GH22 but also on GH21, GH20 and GH19. Guardant argued that “Factsheet GH20 confirms that the MSK-DDM assay is based on “deep sequencing (...). This information is also clear in GH19, GH21, and GH22” (application para. 205, emphasis added). 125. Feature 1.3 is sufficiently evidenced by GH21, where it reads “coverage at any position within the target regions is limited to 200,000” and coverage is controlled only when a depth ˃˃50,000x is reached. Guardant expressly referred to GH21 in its application. There, Guardant explained that “this point can also be considered by working backwards from the sequencing stage” (emphasis added) and referred inter alia to GH22. Given this, it is not clear why Guardant would only have assumed that feature 1.3 was fulfilled after becoming aware of GH22. 126. Regarding features 1.4 and 1.6, Guardant argued in the Application that “workflow (...) of GH22 shows that the enrichment via hybridization capture precedes the sequencing steps. (...), Amplification is performed before capture (...) and again afterwards” (application, p. 72). As GH21 also describes post-capture amplification (For example: “The recommended number of post-capture amplification PCR cycles, sequencing depths per sample, run composition recommendations, and sequencer loading dilutions can be found in the library preparation IFU document (SG-05826) which accompanies the Product” (GH21, p. 13, 7.2.1 Data Generation, sentence 2), and post capture amplification necessarily involves prior enrichment, it follows that the realisation of features 1.4 and 1.6 was already apparent from GH21. 127. Since GH21 refers to GH22, obtaining this document would indeed have been part of the due diligence process in preparation for the proceedings. However, as Guardant already had all the information necessary to establish patent infringement based on GH21, the question of when Guardant knew, or ought to have known, about the infringement does not depend on its receipt of GH22. It is only a relevant factor for answering the question whether the delay after knowledge of the infringement is reasonable. However, as Guardant did not credibly demonstrate that it received or could have received GH22 only on 11 July 2025, it is also irrelevant in this context. **No knowledge of realisation of features before May 2025** 128. Contrary to Sophia's view, it cannot be assumed that Guardant knew or should have known, before May 2025, that the features of claim 1 of the patent had been realised by the attacked embodiment. 129. Although GH21 was published online as early as October 2024, Guardant could not be expected to be aware of it earlier than May 2025. As a general rule, the patent holder is not obliged to monitor the market. However, if the patent holder becomes aware of specific circumstances that suggest patent infringement, the patent holder is expected to investigate the matter with due diligence so that the patent holder can take action against all infringers as quickly as possible. Sophia did not provide any circumstances suggesting that Guardant had or should have been aware of a patent infringement before 1 May 2025. The fact alone that the attacked embodiments have been available on the European market since April 2024 is not sufficient. 130. A long-standing infringement may indicate that the patent holder has turned a blind eye to the infringement. However, the Paris Local Division rightly held that information available prior to May 2025 is either too general in scope (i.e. commercial documents that do not contain technical information) or relates to activities outside Europe (particularly in the USA). 131. The fact that Guardant's employees were attending Sophia's online seminars does not justify the assessment that Guardant should have known the infringement of claim 1 of the patent at issue. If the applicant is a legal entity, the decisive factor for assessment of urgency is when its authorised representative body or an individual, who by virtue of its position within the legal entity is capable of pursuing the infringement or reporting it, internally becomes aware of it. For example, this person could be an employee of the legal department, a senior member of the sales department or a mystery shopper engaged specifically to investigate intellectual property rights infringements. As witnessed by Guardant's Vice President (GH40) none of the employees attending Sophia's online seminars were involved in evaluating potential patent infringement claims and none of them had an obligation to forward information of the activities of other companies to the decision makers. 132. The fact that the email address of Guardant's representatives was used to access the GH23 webinar for the first time on 18 April 2025 shows that Guardant was investigating the attacked embodiment in April 2025. But Sophia itself does not submit that the GH23 webinar showed all features of claim 1 of the patent at issue. **Knowledge of, or opportunity to gain knowledge of activities in Europe** 133. As stated above, if the patent holder becomes aware of specific circumstances that suggest patent infringement, the patent holder is expected to investigate the matter with due diligence so that it can take action against all infringers as quickly as possible. It follows that as soon as Guardant was aware of infringing activities of Sophia in the UK, Guardant was expected to investigate whether there were the same infringing acts also in UPC territory, Switzerland and Spain. 134. Guardant asserted it did investigate Sophia's activities in the UPC territory, Switzerland, Spain and Norway, but did not specify when it first became aware of Sophia's infringing activities in these territories in the CFI proceedings. Guardant submitted in the CFI proceedings that during the preparation of the UK action, it noticed that the Sophia companies were also active in the EU and began investigating the situation in Europe beyond the UK and that only during this additional research Sophia's activities within the UPC territory became clear. Guardant submits in the appeal proceedings that it did not start its investigations regarding the UPC countries before May 2025. 135. As it must be assumed that Guardant had knowledge of the alleged realisation of the claims of the patent at issue by the attacked embodiments by 1 May 2025 (see para. 116-121 above), this is when Guardant should have started to investigate Sophia's activities in the UPC territory, Spain and Switzerland. 136. If Guardant launched its investigation on 1 May, it is difficult to understand why it took until 13 August 2025 ‒ a period of more than three months ‒ to find sufficient evidence showing that Sophia was selling the infringing products within the UPC's jurisdiction, and that all the Sophia companies were responsible. The same applies even if Guardant launched its investigations on 31 May 2025. 137. There was sufficient publicly available information showing that all the respondents were offering the attacked embodiments. For example, in its warning letter, Guardant identified a financial report (SG41) from Sophia as evidence of sales to UK customers. According to this financial report Sophia has signed new customers in Belgium and France (SG41, page 1). Another financial report (SG50), available at the same time and in the same place, details how the University of Heidelberg recently completed implementation of the attacked test. Furthermore, it states: “These institutions will ramp up their usage in Q4 2024 and into 2025”. 138. In its application for the liability of respondent 1 Guardant refers to the company's international website, for which respondent 1 is responsible and which is directed at customers worldwide, including those in the UPC member states as well as in Spain, Switzerland, Poland and Norway. Regarding infringing acts of Respondent 2, Guardant refers to the subpage of the website where Respondent 2 is explicitly mentioned in the “Contact us” section and to the introduction of privacy policy (GH24), which is available from this website. Furthermore, Guardant relies on the fact that Respondent 2 is named as authorized representative of Respondent 1 and as the importer of products manufactured by Respondent 1 in the European Database on Medical devices. Regarding Respondent 3 Guardant also refers to GH24 in its application, which is available from the company's website. Regarding Respondent 4 Guardant relies in the Application on the business register entry. 139. Since all information ultimately relied upon by Guardant consisted of publicly available corporate records, website information and publicly accessible commercial material, no substantial investigative measures were required. The Court therefore considers that a diligent patent holder could reasonably have completed this inquiry within a short period of time, which in the circumstances of the present case would not have exceeded two weeks. In conclusion, with due diligence Guardant could have been aware of the patent infringement within UPC territory, Spain and Switzerland by all of the Sophia companies by 15 May 2025 at the latest. **Technical analysis** 140. A patent holder is not required to take every risk when filing an application for provisional measures. Instead, it is entitled to prepare for every possible procedural scenario that may arise, so that it is equipped to respond successfully to any response or defence from the opposing party and present the documents it considers necessary for the success of such an application. In principle, the patent holder cannot be required to conduct further investigations or obtain the necessary documents after having filed the Application. Thus, the patent holder cannot generally be accused of carrying out additional tests or a technical analysis as a precaution before it lodges an application for preliminary measures (see CoA, 14 February 2025, UPC CoA 382/2024, Abbott v Sibio, para. 149). 141. It is for the patent holder who files an application for provisional measures to explain, in a substantiated manner, where applicable, which further analysis or tests were required, for which claim(s) or claim features it was necessary to conduct them, how long this took and why this time was reasonable. 142. Against this background, Guardant's assertion, not evidenced by any documentary evidence or witness statements, that technical analysis of Sophia's test between May and July 2025 was not straightforward and took several weeks due to the complexity of the patent's technology and the lack of public technical information on how MSK-DDM works, is insufficient. Guardant has not specified when the technical analysis began, which patents they were essential for, how long they took and why this time period was reasonable. 143. Guardant's argument that it was not readily apparent whether MSK-DDM used non-unique tagging (feature 1.2 of EP’073: a converting initial starting genetic material into the tagged parent polynucleotides using non-unique barcode oligonucleotides, wherein converting comprises enzymatic ligation) is not relevant. This would have justified the delay for filing the Application with regard to EP ́073, but not for the patent at issue. **No reasonable delay** 144. That in contrast Guardant's knowledge of Sophia's commercial activities within the UPC territory, Spain and Switzerland sufficient to initiate patent infringement actions in relation to ”the attacked embodiment was not available until 13 th August 2025” (Witness Statement GH40) and therefore nearly three months after Guardant could have acquired knowledge of the infringing activities of Sophia in these territories, shows that the enforcement of Guardant's rights was not urgent for Guardant. 145. There are no circumstances that would justify Guardant's delay in submitting the Application until 29 August 2025. As Guardant explained at the oral hearing before the Court of Appeal, Guardant needed two weeks for drafting the Application. This means that Guardant started with drafting not before 13 August 2025. As considered above, the delay between the date of awareness of infringing acts within the UPC territory, Spain and Switzerland on 15 May 2025 and 13 August 2025 has not been accounted for by Guardant. 146. The fact that the patent is complex, as Sophia admits, cannot explain why the Application was only lodged on 29 August 2025. As stated above (para. 142), as Guardant has not specified when the alleged technical analysis and/or tests began, which patents they were essential for, how long they took and why that was a reasonable time, the alleged tests cannot justify the delay. Indeed, even if these took ‘several weeks’ as Guardant submits, which must be taken to mean no longer than a month, a delay of still two months lacks explanation. Similarly, Guardant's explanation at the oral hearing before the Court of Appeal that Guardant had to consult technical experts lacks substantiation in this respect, as Guardant did not specify which features of which patent required examination or how long this took and why. 147. The late response to Guardant's warning letter cannot justify such a delay either. As a general rule, the patent holder may wait a reasonable time for a response to a warning letter before drafting and lodging an application for provisional measures. However, as the warning letter did not refer to the patent at issue, there was no reason for Guardant to wait for Sophia's response on 20 June 2025 before drafting an application for provisional measures with regard to the patent at issue. **E. Conclusion** 148. As there is no urgency, Guardant's appeal is unsuccessful. The further grounds for invalidity put forward by Sophia are therefore irrelevant, as are the questions of whether the patent at issue is infringed and whether preliminary measures are necessary. **F. Permission of allegedly late filed arguments** 149. Insofar as Guardant objects to Sophia's alleged new arguments regarding invalidity on the grounds that they were raised too late, no decision is required, as these arguments are not decisive or do not lead to the success of Guardant's appeal. The same applies for some passages in the Statement of response to the cross-appeal, which Sophia objects to as late filed. **G. Cross-Appeal** 150. Pursuant to R. 220.1 RoP, only a party adversely affected by a decision may lodge an appeal. The same applies to a cross-appeal pursuant to R. 237 RoP. A cross-appeal is inadmissible if the only purpose of the cross-appeal is to change (a certain part of) the reasoning of the Court in First Instance, which in its result is in favour of the party filing the cross-appeal (CoA, 17 April 2026, UPC CoA 901/2025, Abbott v Sinocare et al., para. 209). 151. In accordance with Sophia's request in first instance, the CFI dismissed Guardant's claims. Guardant filed an appeal against this decision. Whether the requirements of urgency have been fulfilled is part of this appeal (R. 211.4 RoP). The same issue cannot form the subject of the cross-appeal. Therefore, the cross-appeal should be declared inadmissible. 152. Insofar as the CFI did not order Guardant to bear the costs incurred by Sophia, Sophia is adversely affected. But this does not mean that the cross-appeal is admissible. According to R. 242.1 RoP the Court of Appeal shall either reject the appeal or set the decision or order aside totally or in part substituting its own decision or order, including an order for costs both in respect of the proceedings at first instance and on appeal. This means that the Court of Appeal decides on costs on its own motion. This even applies where the CFI did not decide on costs at all. 153. The inadmissibility of the cross-appeal does not mean that the submissions in the cross-appeal have to be disregarded. The Court treated it as part of Sophia's statement of response. **H. Value of the appeal and cross-appeal** 154. The CFI set the value of the application for provisional measures at EUR 6,000,000. As the impugned order has been appealed only insofar as it relates to one of four patents and the interim award of costs, the value of the appeal is lower. 155. Contrary to Sophia's opinion, the value of the appeal and the cross-appeal is not the same as the value of the original Application. The various patents in question relate to different subject matters and are all economically significant to Guardant. Sophia's modification of the contested embodiment may mean that the product no longer falls within the scope of protection of one patent, while still falling within the scope of protection of the others. 156. Guardant assesses the value of the patent at issue of EUR 1,500,000. The Court of Appeal agrees. 157. However, it must not be forgotten that the appeal also concerns the interim award of costs for Sophia (EUR 400,000), which also concerns the other patents. Insofar as it concerns the other patents, the value is EUR 300,000. This justifies setting the value at EUR 1,800,000. **I. Court fees** 158. Based on this value, the court fee for the appeal and the cross-appeal is EUR 33,500 (14,600 + 18,900 value-based fee). Guardant and Sophia paid EUR 71,400. This justifies a reimbursement for each of EUR 37,900. **J. Interim award of costs ordered by the CFI in favour of Sophia** **Obligation to pay interim awards of costs** 159. The CFI ordered Guardant to pay Sophia “interim costs of the proceedings amounting to 400,000 Euros”. 160. An interim award of costs may also be ordered in favour of the defendant in proceedings for provisional measures (CoA, 28 November 2025, UPC CoA 317/2025, Barco v Yealink, para. 96). 161. Guardant claims unsuccessfully that the interim award is based on a manifest error because the CFI did not issue a cost order and Sophia did not file an interim costs award. The CFI established in the impugned order that both parties requested reimbursement of the costs amounting to EUR 600,000 (para. 154). This is accurate, as Sophia requested in the Objection (page 93) that Guardant be ordered to pay an interim award of costs in the amount of EUR 600,000. 162. It is true that an explicit order for costs is missing in the impugned order. But this does not influence the order granting an interim costs award. 163. Guardant argues that Sophia had early opportunities to provide their non-infringement argument for the patent EP’533. It was cited in the UK warning letter. Sophia's failure to respond appropriately to Guardant's correspondence caused unnecessary costs that should be borne by Sophia. 164. The Court of Appeal does not agree. As a general rule, the unsuccessful party must bear the reasonable and proportionate legal costs and other expenses incurred by the successful party (Art. 69(1) UPCA). Exceptions apply where equity requires a different allocation of costs, in particular where a party succeeds only in part or in exceptional circumstances (Art. 69(2) UPCA) and where the party has caused unnecessary costs to the Court or another party (Art. 69(3) UPCA). As a rule, the withdrawal of an action or application means that the defendant must be considered the successful party. Equity does not provide otherwise in cases where the claimant applies to withdraw the application for provisional measures or the action after the defendant, for the first time in the proceedings and not in response of the warning letter, raises an argument against infringement. In principle, a person who has received an unjustified warning is under no obligation to provide clarification why it will not comply with the request. **Amount of the cost award** 165. Guardant successfully challenges the amount of the interim costs awarded by the Paris Local Division. Guardant is only obliged to pay 50% of the applicable ceiling (EUR 600,000) set by the decision of the Administrative Committee of 24 April 2023 for a case with a value estimated at 6 million euros, thus EUR 300,000. 166. To what extent the successful party is entitled to a reimbursement of representation costs is to be determined in the proceedings for a cost decision pursuant to Chapter 5 of Part 1 RoP. In general, the Court may assume that the successful party will be entitled to 50% of the applicable ceiling and may order reimbursement of the amount by means of an interim award, unless there are clear indications that the successful party in fact incurred fewer representation costs, or that 50% of the applicable ceiling is more than what would be reasonable or proportionate in the particular circumstances of the case. At the same time, as a rule, the Court cannot assume that the successful party is entitled to more than 50% of the applicable ceiling before the conclusion of the cost proceedings according to Chapter 5 of Part 1 RoP. An exception may apply if parties have submitted and discussed cost specifications during the proceedings or have agreed on the costs to be reimbursed (CoA, 25 November 2025, UPC CoA 464/2024, Meril vs Edwards, para. 203). 167. In this case, there is no reason to depart from the rule. The mere fact that both parties have each applied for an interim costs award amounting to 100% of the applicable ceiling does not justify the assumption that they have agreed on this amount to be reimbursed. It is also not sufficient that Guardant did not challenge the amount of Sophia's interim costs award request in the CFI proceedings. 168. It is not objectionable that the CFI did not consider the withdrawal of the application relating to EP’533 to be a ground for reducing the interim costs award. As the CFI rightly held, Sophia had to examine its defence and to draft the Objection also with regard to EP ́533. 169. Sophia filed a submission on 9 December 2025 that did not comply with the CFI's order at the case management meeting and was disregarded. Sophia's costs of preparing this invalid submission was unnecessary and must be borne by Sophia. But this is to be determined in the proceedings for a cost decision pursuant to Chapter 5 of Part 1 RoP and does not affect the assumption that the successful party is entitled to 50% of the applicable ceiling. **Time period interim award** 170. The Court of Appeal considers a period of two weeks for payment of the interim award to be reasonable. **K. Interim award of costs for the appeal proceedings** 171. As the same principles apply to the interim costs award in the appeal proceedings, it is justified to grant 50% (EUR 100,000) of the applicable ceiling (EUR 200,000) for a case with a value of EUR 1,800,000 (see para. 157). **L. Costs** 172. With regard to the cross-appeal, Sophia is the unsuccessful party. According to Art. 69(1) UPCA, as a general rule, the unsuccessful party has to bear reasonable and proportionate legal costs and other expenses incurred by the successful party, unless equity requires otherwise. Here equity requires that Sophia must only bear the court fees caused by the cross-appeal. Since the assessment of whether the requirements of urgency and necessity have been fulfilled is a necessary part of the main appeal, it is justified that Guardant bears the costs caused by the statements concerning the cross-appeal. 173. The appeal is largely unsuccessful. As Guardant's successful claim was relatively minor and did not cause further costs, Guardant shall be ordered to pay the costs of the proceedings at first instance and on appeal (see CoA, 30 April 2025, UPC CoA 768/2024, Insulet v EOFlow, para. 135). **ORDER** The Court of Appeal I. sets aside point 3 of the order of the Paris Local Division of 23 January 2026, UPC CFI 808/2025, in part, and sets the amount of the interim award granted by the Paris Local Division to EUR 300,000; II. rejects Guardant's further appeal; III. declares Sophia inadmissible in the cross-appeal; IV. orders Guardant to bear the costs and other expenses incurred by Sophia in the CFI proceedings and on appeal, including the application for suspensive effect and the cross-appeal, with the exception of the court fees paid in respect of the cross-appeal; V. orders Guardant to pay to Sophia as an interim costs award for the appeal proceedings EUR 100,000 no later than two weeks after lodging this order; VI. orders the overpaid fees in the amount of EUR 37,900 to be reimbursed to Guardant pursuant to R. 371.4 RoP; VII. orders the overpaid fees in the amount of EUR 37,900 to be reimbursed to Sophia pursuant to R. 371.4 RoP; VIII. rejects any further requests made by Guardant and Sophia; IX. sets the value of the Appeal proceeding to EUR 1,800,000; X. declares this order immediately enforceable. Issued on 2 July 2026 Rian Kalden, presiding judge and legally qualified judge Patricia Rombach, judge-rapporteur and legally qualified judge Ingeborg Simonsson, legally qualified judge Arwed Burrichter, technically qualified judge Andreas Gustafsson, technically qualified judge For the Registry:

Key Holdings

• The requirement of urgency for a preliminary injunction (PI) is strict; patentees must demonstrate diligence and provide substantiated explanations for any delays.
• International jurisdiction for the UPC can be established under Art. 8(1) of the Brussels Regulation when an anchor defendant and another entity are jointly responsible for alleged infringements.
• Added matter and claim interpretation are questions of law, allowing new legal arguments to be raised on appeal, provided they are based on the same facts, though potential for surprise decisions needs careful management.
• A cross-appeal is inadmissible if it challenges only the reasoning of a decision, rather than its operative part, including decisions on costs.
• Knowledge of infringement for urgency purposes can be imputed if, under the circumstances, a reasonable person would have become aware of the alleged infringement.

Tags

• Urgency
• Preliminary Injunction
• Appeal
• Jurisdiction
• Added Matter
• Claim Construction
• Costs
• Cross-Appeal
• Brussels Regulation

Related Cases

• UPC_CoA_19/2026 – Guardant v Sophia
• UPC_CFI_808/2025 – Guardant v Sophia
• UPC_CFI_808/2025 – Guardant Health v Sophia Genetics
• UPC_CoA_49/2026 – Guardant v Sophia