AOAC Gluten Qualitative Validation Guidance-FINAL (July 2023)

Gluten Working Group Comments/Responses (Round 2)

4.1.3 Matrix/POD Study • I would prefer scheme 2 when dealing with different gluten sources (page 24 of the working book) but I do not understand how these different schemes will be mentioned in the document that we have to vote for; without these schemes being discussed I can not vote with "yes" • The document cannot be voted for or against without clarification on the necessary matrix / gluten source combinations (page 24 of the booklet). Proposal #1 is very laborious and it is not clear if this much more extra work will result in much more gain of knowledge (e.g. the response to the gluten sources may be independent of the matrix). Proposal #3 may be difficult to perform, due to different CDCs for each gluten source. Proposal #2 is therefore preferred. As a compromise between proposal #1 and #2, checking of all gluten sources in one matrix for every five claimed matrices / for every claimed matrix category may be favorable. All other matrices will be tested either with the gluten source showing the highest CDC or with wheat (as wheat is by far the most important gluten source). The matrix chosen to test all claimed gluten sources should be one of the more challenging matrices (most highly processed). • We need to choose a scheme for dealing with multiple gluten sources. I'm fine with #2. For #3, we should not require this part: Or if you wanted to claim 4 or 5 matrices, you could do this for the first three, but would then need to do all sources at all levels in the other two. I would not approve scheme #1 o There are no longer multiple options for dealing with multiple gluten sources – the previous version of the document was accidentally posted for a vote, and in the correct version there is only one option given. These comments above were received before that error was discovered. • Tables 1 and 2. The random assignment of gluten sources and different matrices will strongly limit the possible conclusions from this experiment. It will remain unclear, if different results will be due to different gluten sources or different matrices. I suggest the following change: one matrix should be tested will all claimed gluten sources. All other matrices should be tested with one gluten source only, but it should be the same one in all matrices (suggestion: either that gluten source should be wheat due to its predominance of contamination or the gluten source with the highest CDC). The matrix to be tested with all gluten sources should be matrix with the highest level of processing. Sections 4.2.2 and 4.3.3 must be changed accordingly. o There is definite value in seeing all gluten sources in one matrix, but there are not often instances when data is compared across matrices, so the tables have been changed to add all gluten sources to the first, most processed matrix. The remainder are a rotation across gluten sources. In a study with a large number of matrices, it would be useful to have more than one data point for each claimed gluten source. 4.1.6 Robustness Study • I still do believe that this linear regression should not be the only possibility. Was it done before or is it just a theoretically? If it was never used before in this area we should mention this as one possibility but not as the only one. o This appears to be a common methodology for evaluating factorial designs and was recommended by a member of the Stats Committee. We will ask that they review whether this should be the only recommended method of analysis.