 And so for working group three we have Simone Pinero, I'm not doing good with any last names today who's joining us from the US Food and Drug Administration presenting on behalf of the working group that was led by her and Bob Davis from the University of Tennessee Health Science Center. So good evening, we had very interesting discussions last evening and Bob and I sort of wrote down what we thought would be the key points or the key take home messages from these discussions but if I'm in any way not representing them well or if I'm not capturing everything that was discussed I'll ask that the group members feel free to chime in. And these were the group members. So we're tasked with several questions and one of them were to identify key gaps in pharmacosurveillance and by far the main challenge that we discussed were difficulties in case ascertainment or identification of the phenotype. We thought that we needed in order to achieve that we would need standard case definitions. I would be helpful to have a minimum set of variables that could differentiate well cases from non-cases and you'll be helpful to have that to be generalizable to common data models so we can use large data. And at the end of the day we need the ability to dig deeper. We need to be able to validate these cases. We need to be able to figure out a timing of the exposure to try and attempt to establish causality between the potential exposure of a causative exposure and the case. So some next steps that we discussed in regards to that particular topic were to evaluate very different case definitions that are currently being used by other groups. Some examples are listed there at Eurus Car, Redus Car Itch and other SJS projects. And at the end of the day this is going to be an iterative process among researchers and clinicians to arrive at a common definition. But that if we can do that, that would be extremely helpful. The bottom line is that active surveillance with real-time data collection is very different from retrospective collected data with case validation and a set of items that could satisfy both collection efforts would be very helpful. We also talked about capabilities of developing active monitoring in the United States specifically. And there are several challenges. One of them is the fact that our healthcare system is very fragmented. But at the end of the day we may need multiple strategies that include use of both perspective and retrospective data collection for prospective data collection efforts. What we'd need here allow us to have more complete case ascertainment for pharmacogenomic studies. And we discussed a lot, perhaps the use of burn units. And it seems to be a promising approach since that's where many of these patients end up being cared for. And perhaps focusing on a few large areas of cities in the United States may be helpful. Giving the difficulties of doing this nationally, of course, that's the ideal. Use of existing large databases for active surveillance can be promising because of the large numbers. However, there's still several things that need to be accomplished in order for us to get there. Some capabilities are being developed in some databases for pharmacogenomic studies, but we need the ability to identify cases reliably. An ability to conduct case validation blinded, of course, exposure status. We need standardized processes for collecting genetic data across disparate sites and so on. And Enbridge, as we heard yesterday, has had success in shuffling this process. We also talked about challenges in estimating rates of SJS and TN. And that's also an issue of particular importance here for the U.S., but a lot of the discussions here, and I'm curious to hear also from others in this room, one of the issues that we discussed is whether this is truly a priority, whether we need to understand rates. And of course, for events as rare as SJS and TN, over time, may not be the helpful but this is probably very relevant for cost-effectiveness study to estimate burden of disease and to calculate product-specific rates. And if we need to do so, of course, near a complete capture of cases, an ability to identify them and enumerate a population is needed. And if we can develop this capability, we can consider assessing product-specific rates, perhaps race-specific rates, which may be very helpful and informative to us. And as we consider that, it would be important, of course, to restrict this type of study to new users of a drug, since if you include prevalent users of a drug, these people may no longer be at risk for SJS and ROTN if you've been using a drug for years. We also talked about gaps in knowledge in terms of understanding disease progression and also gaps in knowledge in terms of understanding long-term outcomes of SJS and TN. Some strategies that we discussed was to consider use of score 10, which is used in Europe mostly as seven-point score used for prognosis, calculated on day three of the disease. Some of the limitations is that data are not always collected on day three. Such as, for example, lab values, and some characteristics such as lung involvement are not considered in score 10. Another next step is to consider studies to address the range and extent of outcomes and disabilities in these patients. And there's also another of loose items that we discussed yesterday that we thought that would be important to bring up. Large-scale collaborations in U.S. might be simulated by concerted efforts to deposit data into public databases, and some of them discussed with databases such as ClinVar and others. And we also talked about the challenges with current screening recommendations because of the low PPV of these, meaning not everyone that screens positive would actually go on to develop the disease. So if we could come up with ways to improving that, perhaps, talking about if you have a high-capital type plus a series of risk factors in order to increase the predictive value of the screening test may be helpful as well. So this is all I have, and I'll open for the group members in case we forgot anything important or was not accurate enough, of course. Lois. We also discussed a network such as Dylan. Remember, that was at the end. And I don't think you mentioned that. You probably mentioned networks in general, but we thought the group thought it would be useful and some members didn't know about Dylan or how it was formed. So they were looking forward this morning to Jehoffly Nagle's talk. So I think we are more informed now. Thank you. Yes, Maya. Yeah, I just wanted to mention when we talked about the clinical classification, there was not an issue that the clinical classifications are no good. They are very good for prospective cases. The difficulty that we identified was when we thought there's a retrospective setting, how could you first have inclusion criteria to identify potential cases and how could you validate them? Because if you are as strict as with cases that you have everything available prospectively from photographs to biopsy to everything, it's not such a difficulty. I mean, sometimes it's challenging in individual cases, but in principle, the setting is there. But for retrospective cases, it can be quite complicated. And that was one thing. And the other thing was trying to make a link between retrospective data assessment and potential genetic investigation or genomics. You know, how could you have access to the blood of patients if you have the data retrospectively? Are there any means or not? Certainly not for the cells, because you need them in the acute stage, but perhaps for DNA. And we also thought maybe there can be a question whether the patient associations could help to donate blood. We have done that in France and we had about 70 cases. But we went back to the case charts, reviewed everything. And then could include these cases for genetic studies because the patient's donated samples for a delay analysis. So that maybe just in addition was some of the talk. Great, Mark. Yeah, just to add on to that, since our charge was surveillance, we took that in a very broad sense. And this relates to the comments I made at work group one is that it was clear that the criteria that would be used for an active surveillance and early warning system would look different than a back end where you want really. So it's really, you want high sensitivity and you can sacrifice some sensitivity if you're doing an active surveillance program. But for that retrospective, you know, research to learn more about it than you really want to emphasize the specificity. And so one of the keys of defining sort of the minimum of data elements for case definition would be to reflect that those elements may be different depending on whether you're in an active surveillance mode versus more of a study research mode. And so the ability to reconcile both of those perspectives and develop tools that would support both of those perspectives as important. And I just wanted to quickly add to what Maya said in terms of getting having the ability of obtaining additional sample from patients. We talked about yesterday about in some of these databases there are patients with their biospecimens available. But when we're talking about SJS and TEN, even if we can identify these cases reliably and retrospectively in these databases, they may not be the ones that have the biospecimens available. So there's still a lot of patients without these specimens available. So the ability to go back and be able to obtain these additional samples in these patients becomes very important in these spare diseases. Any questions or feedback from people who weren't in that particular working group?