The Limits of G 2/21: T 0655/24 on Post-Published Evidence and Improved Effects
Board 3.3.04 rejects recent case law allowing post-published evidence to prove an improved effect, holding that improvements must be credible from the application as filed.
Picture defending a European patent application where your claimed antibody variant shares its core mutations with the closest prior art, but adds one more substitution. You have post-published data proving this extra mutation delivers a significantly improved reduction in receptor binding. The question is: can you rely on that improvement for inventive step if the original application only showed the baseline effect?
The triple mutation dilemma
In T 0655/24, the patent (EP 2 943 507) claimed a protein with a human IgG1 Fc region comprising specific amino acid substitutions: L234F, L235E, and D265A (the "FEA" variant). The claims required that the Fc region does not bind to any Fcγ receptors while maintaining a plasma clearance rate comparable to a wild-type protein. The closest prior art, document D10, disclosed the "FE" variant (L234F, L235E) which already exhibited reduced effector function.
The proprietor sought to establish an inventive step by arguing that the addition of the D265A mutation resulted in an improved effect—specifically, a further reduction in T-cell activation as measured by CD69 expression. To support this, they submitted post-published evidence (D25 and D26).
Rejecting the implicit improvement doctrine
The Technical Board of Appeal had to determine whether this improved effect could be taken into account under the framework of G 2/21. The proprietor's position was supported by a recent line of case law (including T 1989/19, T 2716/19, and T 0840/22) which held that if a baseline technical effect is derivable from the application as filed, an improvement of that effect is also implicitly derivable and can therefore be supported by post-published evidence.
Board 3.3.04 firmly rejected this approach. The panel held that an improved effect not credibly achieved at the filing date cannot be considered "encompassed by the technical teaching and embodied by the same originally disclosed invention" (reasons 58). The board noted that allowing an applicant to rely solely on post-published evidence to establish an improvement over an effect already known in the prior art contradicts the "first to file" principle and the requirement that an invention must be completed at the filing date (reasons 63–64).
Salvation in the original data
Despite this strict legal holding, the patent survived. The board turned back to the application as filed to see if the improvement was already credible without the post-published data. Figure 3A of the application directly compared the prior art FE variant against the claimed FEA variant across a large concentration range. The board observed a consistent difference in CD69 expression between the two variants, rendering it credible from the original filing that the FEA variant achieved the improved effect of decreased CD69 release (reasons 74).
Because the improvement was credibly disclosed in the application itself, the objective technical problem could be formulated as the provision of a protein having an improved Fc region. Since the prior art provided no guidance that combining the untested FE variant with D265A would yield this improvement while retaining wild-type plasma clearance, the board acknowledged an inventive step.
Practical implications
This decision creates a clear doctrinal split on how G 2/21 applies to improved effects. Practitioners can no longer safely assume that establishing a baseline effect in the application opens the door to proving an improvement later. When drafting a European patent application, if an embodiment is intended to provide an improvement over the prior art, that improvement must be credible from the application itself. Where possible, include direct comparative data against likely closest prior art variants in the original specification. Relying on post-published evidence to elevate a known baseline effect into an inventive improvement is now a highly risky strategy before the EPO.