As early as T 35/85 it was stated by the the board that an applicant or patentee may discharge his onus of proof by voluntarily submitting comparative tests with newly prepared variants of the closest state of the art identifying the features common with the invention, in order to have a variant lying closer to the invention so that the advantageous effect attributable to the distinguishing feature is thereby more clearly demonstrated (T 40/89, T 191/97, T 496/02, T 765/15, T 1323/17).
According to the established case law, an unexpected effect (advantageous effect or feature) demonstrated in a comparative test can be taken as an indication of inventive step (T 181/82, OJ 1984, 401). In T 197/86 (OJ 1989, 371) the board supplemented the principles laid down in T 181/82, according to which, where comparative tests were submitted as evidence of an unexpected effect, there had to be the closest possible structural approximation in a comparable type of use to the subject-matter claimed. It stated that if comparative tests are chosen to demonstrate an inventive step on the basis of an improved effect over a claimed area, the nature of the comparison with the closest state of the art must be such that the alleged advantage or effect is convincingly shown to have its origin in the distinguishing feature of the invention compared with the closest state of the art (see also T 234/03, T 568/11, T 1457/13, T 1521/13; T 1401/14; T 710/16, T 990/17, T 2406/18, T 816/16)
For this purpose it might be necessary to modify the elements of comparison so that they differed only by such a distinguishing feature (see e.g. T 197/86, T 292/92, T 819/96, T 369/02, T 2043/09, T 183/12, T 710/16, T 990/17).
To be of relevance in demonstrating that a technical improvement is achieved in comparison with the closest state of the art, any comparative test presented must be reproducible on the basis of the information thus provided, thereby rendering the results of such tests directly verifiable (T 494/99, T 234/03, T 236/09, T 1962/12, T 383/13). This requirement implies, in particular, that the procedure for performing the test relies on quantitative information enabling the person skilled in the art to reproduce it reliably and validly (T 234/03, T 236/09, T 383/13, T 1962/12, T 532/14, T 795/14). Vague and imprecise operating instructions render the test inappropriate and thus irrelevant (T 234/03, T 236/09, T 383/13, T 1962/12, T 795/14).
In T 2579/11 the patent proprietor had not divulged the details of the test protocol applied, citing commercial reasons. The board found that the test results were neither verifiable nor plausible and therefore insufficient to establish that the problem had been solved successfully.
In T 702/99 the board stated that in cases concerning products such as cosmetics, in which applicants or patentees seek to establish that their inventions have an improved "feel" over the prior art, or opponents seek to deny such an improved "feel", it is common for one or more parties to file evidence of comparative tests conducted by a number of persons. It is essential for such tests to be made under conditions which ensure maximum objectivity on the part of those conducting the tests and who may be required at a later date to give evidence in proceedings. It is always desirable that such tests can be shown to be "blind" and that they have been conducted in the strictest conditions; that the testers have had no part in the making of the claimed invention or research leading up to the invention or the patenting procedure. See also T 275/11, T 1962/12, T 165/14 T 2304/16.
In T 172/90 the comparative examples produced did not constitute suitable evidence of inventive step. The board said that the products adduced as a basis of comparison were commercially available and had evidently been selected at random. Technical progress shown in comparison with products of this kind could not be a substitute for the demonstration of inventive step with regard to the closest prior art (referring to T 164/83, OJ 1987, 149). See also T 730/96.
In T 390/88 the board stated that where an alleged invention was prima facie obvious having regard to the prior art, it was however sometimes possible to prove inventiveness by comparative tests showing a significant improvement over the closest prior art. That situation was to be contrasted with other cases, in which it was not prima facie obvious to make the claimed compounds at all, and therefore comparative tests were not essential to establish inventiveness. See also T 656/91, T 60/95, T 930/99.
In T 2319/14 the board disagreed with the respondent's argument that comparative tests always had to be carried out using the closest prior art and that intrinsic comparative tests were not allowed. It was true that, according to the boards' established case law, tests comparing the invention with the prior art had to be conducted in such a way that any effect could be attributed to the distinguishing feature. However, it was also permitted – and might even be necessary – to modify prior-art embodiments in line with the invention to such a degree that the only remaining difference was the feature distinguishing the claim.
In T 1323/17 what counted in the board's opinion was not only whether a causal link between a distinguishing feature over the closest prior art and an effect was demonstrated within the framework of a comparative test submitted by the applicant or patentee. It also counted whether the variant of the closest prior art selected as a reference (or comparative) example for the comparative test was representative of the closest prior art, in the sense that the effect shown to be caused by the distinguishing feature in the context of the comparative test could also be expected to take place within the framework of the closest prior art despite the existence of differences vis-à-vis the reference example of the comparative test.