 Good afternoon. I'm presenting on behalf of Simone Pinheiro and myself. And our topic is pharmacosurveillance for SGS and TN in the US. Now, as far as we know, and we would be happy to be corrected, this is mostly carried out at the US Food and Drug Administration. So our presentation will focus largely on what takes place at the FDA. The presentation will take the following format. First, I'll list the tools that we currently use. I'll describe them, including their strengths and limitations. I'll summarize and try to identify gaps in surveillance and then close with suggestions for possible improvements. Now, our pharmacosurveillance tools can be divided into pharmacovigilance and pharmacoepidemiology. For pharmacovigilance, our main tool is the FDA Adverse Event Reporting System, or FAIRS. We sometimes use data mining. We also use the medical literature and VGBase. Now, the pharmacoepidemiology tools, we have them at our disposal, the National Electronic Injury Surveillance System, Co-operative Adverse Drug Event Surveillance. Now, that's a mouthful. So happily, we can abbreviate it to NISCADES. We also do pharmacoepidemiology studies, mainly in databases. And we have another subset of pharmacoepidemiology studies called Sentinel, or its pilot program, the Mini-Sentinel Program. Now, I'm going to spend a fair amount of time on FAIRS, because this is the main pharmacovigilance tool at our disposal at FDA. FAIRS is a computerized database which houses spontaneously reported adverse events associated with human drugs and therapeutic biologics. At the moment, we have more than 10 million reports in FAIRS, and we have been collecting data since 1969. For the last two years, we have added approximately a million reports a year. This cartoon demonstrates how case reports get interferes. Patients, consumers, and health care practitioners on a strictly voluntary basis report either directly to the agency or to the manufacturers of the drug products. The manufacturers are then required by law to report them to the FDA, particularly the most serious ones. And the vast majority of our reports come via the manufacturers. Now, the strengths of FAIRS are that it's simple, it's relatively inexpensive, and it's very good at detecting rare, serious adverse events with a short latency period, such as SGSTN. It's hard to detect SGSTN in clinical trials because of the relatively small numbers of subjects that are enrolled in clinical trials. The other strength of FAIRS is that it is inclusive. All ages, all populations, all marketed drugs and therapeutic biologics. Like all passive surveillance systems, the main limitation of FAIRS is that there is substantial under-reporting. So we have no clear numerator, and we do not have a denominator, so we cannot use these data for incidence calculations. Another big limitation is that information is often missing. Reporting also varies with time and with certain activities in the society, such as publicity and litigation. Now, I thought you might be interested to know that in the last five years, only 0.1% of our reports in FAIRS have been for SGS or TN. Now, how do we detect a signal for SGS and TN? The safety evaluators in our office, they are the officers who are tasked with monitoring certain groups of drugs. And they are mostly pharmacists, and they receive alerts from FAIRS on a daily or weekly basis of the severe or serious adverse events that have been reported in association with their drugs, including SGS or TN. I have data mining here in parentheses because we seldom actually use data mining for SGS and TN since a single case is enough to constitute a signal. We have alerts in the medical literature for the drugs on SGS TN, and we get information from other regulatory authorities around the world, and we also use Vigibase. We find Vigibase is especially useful when a drug is recently introduced to the market or is about to come onto the market in the US, but has already been in use elsewhere in the world. Now, I told you that there was a lot of missing information in FAIRS data or in FAIRS case reports. So I'm going to just illustrate with a couple of samples. The first is a report from a nurse practitioner who reported that a female patient of unknown age developed SGS on an unknown date while on drug A. Concomitant medications and comorbidities were unknown, as was the outcome. Follow-up was attempted, but was not successful. So what do we do with a case like this? Well, we did not include it in our case series because we judged it to be unaccessible because there was just not enough information. However, if this patient had come in as an alert into a safety evaluator's inbox, they would have had to do a complete search of the literature and other databases. It would have constituted a signal, although it would not eventually have made it into a case series for review. This is another example, a 52-year-old man who was on drug X for diabetes mellitus. After nine months, he was still not well controlled, so drug Y was added. And 13 days later, he developed a generalized erythematous rash with bilateral conjunctival hyperemia. He visited a dermatologist who diagnosed that he was having SJS and admitted him to hospital, where all his drugs were discontinued and he was treated with systemic steroids and seen by ophthalmology. He was discharged after one month in hospital with all his symptoms resolved. Now, we assess the diagnosis as being likely because it was made by a dermatologist, and we assess the causality of the drug, of drug Y as probable since it occurred in a close temporal association with drug Y, whereas he had been on drug X for a long period of time without any problems. So this is how we assess diagnoses with the limited information that we have in fares. We say that the diagnosis is likely if it's made by a dermatologist, if there is a good clinical description, especially if there's a record of the percentage of the body surface area that is involved in the blistering and detachment. We accept the diagnosis. If the patient was nursed in ICU or a burn unit, if the patient had a biopsy and the biopsy confirmed that the diagnosis was SJS or TN, but we say that it is less likely but still possible if the diagnosis was made by somebody who was not a dermatologist, and there is no supporting information as above. Now, for causality assessment, we use a modification of the WHO Uppsala Monitoring Center in criteria. We say that the causality is probable if there is a reasonable temporal association, an absence of confounding factors such as concomitant drugs or other alternative explanations. If there is a positive de-challenge and re-challenge information, but if there is only a reasonable temporal association but there are confounding factors, then we say that the causality is possible. Now, people have asked, especially our colleagues in academia over the years, why don't we use the all-den criteria? The answer is that we simply do not have enough detail in our reports and fares. These are not prospectively collected data, they're not collected by a registry, so the data is skimpy at best. But we do use similar elements, we use temporal association, we use de-challenge and re-challenge information, we consider whether they're confounding factors and so on. But the differences, as I said in the detail and the assigning of numeric scores to the different elements, also you heard from Dr. Muckin-Haupt that one element of causality is the notoriety of the drug. Well, at FDA, we are often dealing with drugs that are not previously known, we are the first to make the discovery, so to speak, so we cannot use that element in our assessment. Now, I turn to NICE-CADES. NICE-CADES is a collaboration between the Consumer Product Safety Commission, the CDC and FDA. It's an active surveillance for adverse drug events that are treated in emergency departments in the United States. It's based on a national probability sample of about 60 US hospitals, so it's nationally representative. These hospitals have to have a minimum of six beds and a 24-hour emergency department. An adverse drug event in NICE-CADES is defined as an emergency department visit for a condition that the treating physician explicitly attributes to the therapeutic use of a drug. Now, this cartoon demonstrates how NICE-CADES data are collected. The patient visits the emergency department, sees the clinician, gets some tests done and gets a diagnosis and is treated and either sent home or admitted to the hospital. And all of this is recorded in the notes. Within 24 to 48 hours, trained abstractors from NICE who are usually nurses, they come along and they go through all the notes of the previous 24 or 48 hours. They select the ones that have a diagnosis of an ADE and they abstract pertinent information from the notes onto a computerized form, including the verbatim diagnosis that the EDE clinician made. And these forms are transmitted electronically to the Consumer Product Safety Commission where data cleanup is carried out and then the databases transferred again electronically to the CDC where further checks and validation of the data are carried out. And importantly, based on the verbatim diagnosis from the EDE clinician, MedRA codes are assigned. This is just a screenshot of the computerized form that the NICE abstractors use in the EDE to transfer the information. It's based on our FDA MedWatch form. So it's a similar sort of information. Now, when we're searching for SJS-TEN cases in NICE Cades, we start with the PT and then we move up to the HLT which is Buller Skin Conditions. We sometimes have to use the HLT because we get more cases that way. You find that it might have been coded as a Buller's Condition, but something in the narrative will tell you the patient was admitted to a bullen unit and then the penny will drop that, okay, this probably was a case of TEN. Now, the strengths of the NICE Cades are that it's nationally representative so we can calculate incidence rates from this. It also can be used as an additional source of cases to supplement FAIRs when we're building case series. The diagnosis is made by the actual emergency department physician so that it's better than just the ICD-9 codes. The limitations, however, are that the diagnosis is not confirmed. There is also a lag time of about 15 months for the data to be updated from one year to the previous year. So the lag time ranges from about 15 months to two years because we just got, for example, the data from 2013. NICE Cades does not capture all cases of SJS-TEN. Doesn't capture cases that are not caused by drugs. It doesn't capture cases that occur when the patient is already in hospital for something else and it doesn't capture cases that didn't make it to the hospital. Now we move along to the pharmacotherapy studies. Prospective studies are best but this has proven very difficult in our setting in the United States. Partly because SJS and TEN is so rare and partly because our healthcare system is so fragmented. For a cohort study, you would require very large numbers of enrolled subjects which makes it prohibitively expensive and as far as we are aware, we're not aware of any studies that have been done like that. So prospective studies are what we have been doing or what have been carried out in the U.S. Sorry, retrospective studies are what have been carried out in the U.S. And these are carried out on administrative claims data and more recently electronic medical records. The strengths are that it's a real world setting but unfortunately as you have heard from multiple people today, the ICD-9 codes for SJS and TEN, even the ones that are more recently introduced in 2008 do not validate at all well. They have a very poor positive predictive value of whether a case really will be a case. So these are not, they have not shown to be useful. Sentinel is a program which was launched by the FDA in 2008 in its pilot project, Minicentinel. Sentinel takes a form of active surveillance but it does it on a subset of the administrative claims data and electronic medical records which have specially formatted and standardized so that they're ready to go for studies. Also, they've developed research modules, standardized research modules which can do lots of different types of studies from simple feasibility studies to more complex studies that require adjusting for confounding and that sort of thing. So these are ready to go modules. But again, because for SJS and TEN, the diagnosis is based on ICD-9 codes, we haven't been able to use it successfully for SJS and TEN. So in summary, FAIRS is the main pharmacovigilance tool that we use at FDA. But it's limited by the fact that we have a lot of missing data and there's under-reporting. NISCADES is useful, but BOR could be done especially with confirming the diagnosis. Pharmacoepidemiology studies are limited severely by the fact that we don't have a useful algorithm for detecting the diagnosis and the ICD-9 codes do not validate well. MACE should be in your handouts. I haven't had time to speak about it today, but it stands for Molecular Analysis of Side Effects. And it's a research tool that we have at the FDA and Dr. Keith Burkhart, who should be here, he is a research collaborator with the outside group that's testing this tool. And now I have the following suggestions. I think perhaps we might be able to interest people in doing targeted active surveillance, for example, in intensive care units or burn units, where we would be likely to find cases. It would be the most severe cases, of course, but it would be easy to be sure that the phenotype was correct and that we would get, I think we would get a good yield from looking at these patients. Another source would be to follow up cases identified in NICE-CADES and see what happened to these patients after they were hospitalized. I don't think that this would be unduly burdensome because last year we did an analysis of hospitalizations for SJSNT and NICE-CADES between 2004 and 2009. And in all that time, there were only 60 cases admitted to hospital. Which averages seven a year. So I don't think it would be unduly burdensome. It would, however, take a long time for us to get results from that. It would take another seven or eight years. But it would be a start. Another idea would be to have a network of dermatologists who are interested in studying SJSNT or severe acute ages adverse drug reactions in general, modeled after the Dylan group. I don't know why it is that dermatologists haven't become, you know, haven't become fired up by that sort of thing. But perhaps Dr. Katz or somebody can inspire them. And finally, I'd like to acknowledge the help of all my colleagues in the Office of Surveillance and Epidemiology at the FDA. Particularly those in my division, the divisions of Pharmacovigilance 1 and 2. Thank you. Thank you very much, Dr. LeGrenade. Questions and comments? Yes, please Andrea. Sorry, you have a bit of a hike to get over to it, unless you want to just kind of come to the table. No, you really do need to use microphone because we're recording. Never considered implementing a mandatory reporting system. As our medical practitioners are the first line of timely information, we may find that SJSNT and they're not as rare as we think it is. We have, our mandate at the FDA is a regulatory mandate. And whenever we have tried to ask for mandatory studies, we get some pushback. As if it's primarily for research, then we can't do it. It has to have a regulatory component. So that's our problem. That's our limitation. We have 5,000 signatures of victims for mandatory reporting. I think if you took it to Congress, it might help us better. Right, I was like, it's futile, but, you know, I brought it. Okay. Thank you. Great, thank you. And Munir has a comment, and maybe while he's commenting, others could think if there's any country that you're aware of that has mandatory reporting. That's what I'm going to comment on. So the UK has a voluntary reporting system, but Sweden and Spain have mandatory reporting systems, and the reporting rates are no better in Sweden and Spain compared to the UK, for example. In fact, they're worse sometimes. That's not good news. Just a comment I'll make about the sensitivity of the NEISSK for identifying Stevens-Johnson syndrome. The experience I had with the vaccine safety data link suggested that most physicians were completely unaware of the link between vaccines and well-known vaccine side effects like seizures or ITP after MMR vaccination. And it makes me think that the sensitivity or the ability of most physicians in the emergency room to readily diagnose SGS as a secondary result of Tegretol started the two weeks ago might be very low. I'm wondering what you thought about that. That is a possibility. I think what we find, actually, is that physicians tend to call everything erythema-multiforme, any drug rash, or they have erythema-multiforme. So this is why we use hospitalizations when we're trying to weed out the cases, because if they were hospitalized, then they're more likely to have been SGS rather than just EM. But I take your point, yes. Yes. I wanted to ask, have you observed that reporting certain associations to drugs differs whether the drug is an old drug or a new drug? You showed that you had a list with a high frequency of, for example, carbamazepine-induced cases or somebody showed this morning. And I wonder, because usually new drugs have a higher chance to be reported. And so you have rather an over-reporting with new drugs and old reactions than with the 150s rash after the same drug that everybody knows. Have you seen that? Yes, we have. We have, if the highest reporting is in the first two to three years and the drug is on the market, and then it falls off dramatically after that, especially once it gets into the label, they say, oh, everybody knows about that. So they don't bother to report it anymore. So I wanted to ask about the comment Lois made about potentially doing some kind of surveillance in burn units. That's an interesting idea. And there is a subspecialty of burn medicine and burn unit nurses, et cetera. I wonder, Neil, do you have any experience with that? Yeah, Elizabeth showed the data that we did when we surveyed our units about what was being reported. But I spoke at the American Burn Association annual meeting about two years ago at a symposium on SGSTN or TENS, as they call it. And it was really well attended. I learned a lot about burn units. There are some specialized burn units that are just plastic surgeons who look after burns. Every bed is for a burn and that's it. They often will have three or four states that send patients to them. So they're pretty sick people. Most other burn units, especially pediatric ones where you wouldn't see it as common, might be a bed and a pediatric ICU and the doc. There's one doctor who looks after these people and he's the burn doc and there's a burn unit. It's a unit in the sense that it's one thing, but it's not a unit in the sense of being an ICU or something like that. So it was a really interesting mix. But I have to say the doctors and their support staff and physiotherapists and social workers and all the people who might be part of a burn unit were very, very interested in trying to do a better job in all aspects of looking after SGSTN. And we're quite cooperative too in giving us information about what their therapies were and what investigations, how many people used Alden, how many used SCORE 10 and all these various measures. So I think it's a reasonable thing. It should be nice to collate that and then follow up with reaching out to those burn units in North America, Canada and the US because they do see them. The issue that we found because we have very close relations, we have really Ontario has a population, the province that I live in of about 13 million people. We have the only true burn unit. There's four full-time plastic surgeons who look after burn patients there. And I started relationships with them as soon as they moved to our hospital from another hospital, which was then closed. But what we found was the biggest issue was getting a diagnosis because if somebody was sent to them with TENS, they said it was TENS. And we had to go, no. This one was like in Planetus. This was fixed drug eruption. Could be scabies. I mean, it doesn't matter. And if the outside pathology had been read by a non-dermatopathologist, if it was sent in as TENS, then the answer usually was compatible with TENS. Even though it clearly wasn't toxic epidermal necrolysis ever either clinically or pathologically. So it's nice to think about the pathology helping, but it doesn't always, if it isn't read by somebody who at least knows the strengths or limitations of what they're doing. So I think it's a great idea because I know they are committed and I think they really would like to be involved in a meaningful way. I can confirm that experience to some extent. We are closely collaborating with burn units in Germany and there's about 20 burn units that are larger and they all have patients with SJST and not so many per year, maybe one to five. So it's distributed all over the country. The good thing is they are interested and the burn surgeons want to learn. But they are not the ones to make the diagnosis. They usually overestimate the extent of detachment because it's difficult for them to deal with all the little spots that are around. So that has some impact in the validation, some impact in the treatment, in the supportive care and so forth. And we closely work together because they want us to help and make not only the diagnosis but also help with the medication history because usually the patients are sent to them, not on first hand, but some patients have been in one or two hospitals before because these are referral centers, they don't walk in there. So to obtain a history is not easy and they focus on what is their job, do the correct treatment. So you have to keep in mind, you cannot just take from the burn unit what is there, you have to go back and kind of trace back the history, I think. So I wonder in terms of reporting, I wonder if our FDA colleagues are aware of any surveillance programs related to drugs where there is mandatory surveillance that is actually mandated by the CDC since they're also interested in mandatory reporting for public health reasons. Are there any such programs? Not as far as we are aware. This nice case was a sort of breakthrough surveillance system when it started in 2004. So it sounds like there is room to be pushing the idea of mandatory reporting. If there are groups, obviously, as government employees, we can't lobby and that, but if there are groups that can, hopefully they've heard. So Terry, just related to that, I mean, again, I think we heard clearly from FDA that that's probably not within their purview, but this would be something that presumably would be within the purview of CDC who I don't think we have represented here. So I mean, but in terms of mandatory, at least infectious disease reporting, that's under their... Yeah, no, good point. So maybe Josh, you could come to the microphone and thank you very much. Dr. Luger, not very much appreciate it. Yes, I don't think the FDA would know. So our next speaker is Dr. Josh Denney while his slides are coming up. Who is from Vanderbilt University and he will be talking about the potential for electronic phenotyping in SJSTEN.