WEBVTT 00:00.000 --> 00:00.700 Hi, everybody. 00:00.700 --> 00:03.000 I am Muhammad Ali from Borenger Ingelheim. 00:03.000 --> 00:06.280 I lead the digital development group over there under ClinOps. 00:06.280 --> 00:09.780 And I'm here now today to have a fireside chat with Dr. 00:09.780 --> 00:13.600 Amy Abernathy, the Deputy Chief Commissioner for the FDA and the CIO. 00:13.660 --> 00:15.500 So looking forward to discussion with her. 00:15.500 --> 00:26.800 I'm sure she has wonderful insights, and we would love to get her perspective on many things here regarding biosecurity, medical devices, and the overall digital health space. 00:26.800 --> 00:27.120 Dr. 00:27.120 --> 00:27.760 Abernathy? 00:28.480 --> 00:35.040 Hi, it's great to meet you, and it's great to be here with you and talk about a pretty important topic from my perspective. 00:36.560 --> 00:38.140 Fantastic, fantastic. 00:38.280 --> 00:56.660 So now with the midst of this pandemic kind of ravaging within the United States and pretty much many parts of the world, there has been a huge bits of pressure on enhancing the digital tools and landscape of the ability of types of data that we're collecting on patients and population in real world. 00:56.700 --> 01:16.880 And with increased investments in digital health so far this year, so far hitting 9 billion just in the first half, which has been a record so far to date, where do we see the role of real-world evidence playing in really understanding what to learn about patients and the FDA's perspective around using real-world evidence for this reason? 01:17.510 --> 01:50.320 So first of all, as I think about the pandemic, if I imagine where we are right now and where we're going to need to be in managing this virus and the health of our population, we need to rapidly answer a huge swath of questions right now, which means that while the best way to answer critical questions is through thoughtful experimental designs like randomized control trials, 01:50.320 --> 01:56.840 we don't have enough clinical trial capacity or enough time to use clinical trials for every question sitting in front of us. 01:56.840 --> 02:05.680 And so we're going to need to leverage all the data sources currently available, including data sources that are classically called real-world data sources. 02:05.680 --> 02:16.780 These are electronic health records and claims data, but also include information from social media feeds or from a fitness watch. 02:16.780 --> 02:28.820 So many different data sources that describe a person's health, their activities, what they've experienced, potential adverse effects of interventions or treatments. 02:28.820 --> 02:49.220 We're going to not only need to learn how to use this information and put it to use for critical questions in COVID-19, we're going to need to understand when we can rely on it, when is it adequate and credible, and when do we need to now really direct our clinical trial resources to answer those questions where real-world data can't help us. 02:50.100 --> 02:51.140 That's very helpful. 02:51.140 --> 02:53.260 Now, you mentioned social media. 02:53.260 --> 03:09.940 So what would be some of the steps or examples the FDA has taken to kind of discern what would be a real data source on social media that will be verified versus that of, let's say, bot posting, that that may lead to some understanding which is not true or lead to an incidence for patients which could cause potential harm? 03:10.420 --> 03:22.720 So before I go to how we might deal with social media and bots, maybe what I'll do is back up a couple of steps to think about what we're learning about where real-world data might be put to task. 03:22.720 --> 03:27.700 And that will then help us start to address the questions that you've brought up around social media. 03:27.700 --> 03:38.940 So at FDA, we have been participating in a number of projects inside the agency and outside the agency to utilize real-world data to address our critical questions. 03:39.220 --> 03:51.320 And as we've been doing that, we've been learning really where we can rely on it, where do we need to do additional work in order to allow real-world evidence to be credible, and where can't we rely on it. 03:51.340 --> 03:57.720 And examples are questions such as what's the natural history of COVID-19? 03:57.720 --> 04:02.420 What are the patterns of treatments patients are receiving and how have those patterns changed over time? 04:02.420 --> 04:12.320 What are new problems we're dealing with, such as the emergence of clotting disorders or the challenges of late effects to the heart? 04:12.320 --> 04:27.760 And then how does that inform, for example, what treatments may have differing impacts, both on the safety side as well as the effectiveness of treatments, and how can we think about the role of diagnostics and differing performance of diagnostics? 04:27.760 --> 04:32.720 So those are some of the kinds of questions that real-world data is being used for. 04:32.720 --> 04:37.120 Now you asked specifically about social media feeds and how do we think about that? 04:37.460 --> 04:44.480 Well, practically speaking, you can imagine that we need to take a series of developmental steps and do this quickly. 04:44.480 --> 04:52.420 So first learning how to use electronic health record data more confidently, claims data, et cetera, and then starting to layer in new information. 04:52.420 --> 05:09.920 And an example is we have been watching some pretty interesting work happening with Twitter feeds and how Twitter can help to understand signals that people are communicating about that may deserve additional deep dive attention. 05:09.920 --> 05:18.920 So, for example, one of the areas that really received a lot of attention was around clotting events. 05:18.920 --> 05:32.020 And doctors were noticing clotting events, patients were describing challenges that sounded like clotting events, which then led to increased understanding that maybe clotting was a challenge in the context of COVID-19. 05:32.020 --> 05:40.300 And so we can use, for example, social media to help to pick up signals and then use other data sets to verify those signals. 05:40.300 --> 05:42.360 And that's something we're just starting to think about. 05:42.360 --> 05:48.680 I wouldn't call this FDA-specific work, but rather work FDA is paying close attention to. 05:49.080 --> 05:54.560 The third part of your question was around where might this go awry? 05:54.560 --> 06:05.580 So, for example, how do we need to think about the role of bots and how that may create misinformation that may also be one of the signals? 06:05.900 --> 06:09.300 And, you know, first of all, be attentive, right? 06:09.300 --> 06:19.340 So, recognize that information may be misinformation and may or may not be nefarious in how it has occurred. 06:19.600 --> 06:33.540 Secondly, to do the experimentation to understand what does inadequate information, misinformation, nefarious information look like, and be able to identify that quickly. 06:33.540 --> 06:39.300 So, oftentimes, bots have particular patterns that can be identified. 06:39.600 --> 06:42.200 And then also be transparent. 06:42.200 --> 06:52.340 I mean, one of the challenges with all these new types of data sources is that perhaps you're saying, well, we're interested in learning more about how social media data can help. 06:52.560 --> 07:00.780 And the really quick retort is, oh, but that's not reliable because dot, dot, dot, one of which is bots. 07:00.960 --> 07:17.440 And by being transparent that we are curious about some of these new data sources, but we're also attentive to the potential risks and downsides allows people to know we're paying attention to all the different aspects. 07:17.440 --> 07:22.360 So, this is something that's high in our mind as we start to think about these new types of data. 07:23.100 --> 07:24.280 That's fantastic, Danelle. 07:24.280 --> 07:27.980 You mentioned risks when it comes to these platforms, et cetera. 07:27.980 --> 07:34.340 Now, with the rise of digital health and many of these tech giants really taking a foray into healthcare. 07:34.400 --> 07:41.920 Now, having been within the clinical trial space for almost two decades now, one of the primary steps to get to clinical trials is really informed consent. 07:41.920 --> 07:44.400 Because without informed consent, you really can't get anywhere. 07:44.560 --> 07:51.580 And the informed consent is always for the intended use of the clinical trial that the patient's participating for, for the data they want to collect for in that trial. 07:51.700 --> 08:06.160 Now, with these tech giants, when you have wearables and devices, even your smartphone with the 11 to 15 sensors that are embedded within the actual phone itself, you're continuously capturing data to understand not just overall demographic information, 08:06.160 --> 08:11.900 but you understand behavioral patterns, you understand behavioral impulse control, geographic location, et cetera. 08:11.900 --> 08:29.560 What is the FDA's perspective in terms, or even Amy's perspective, really in this understanding what is the patient's rights in being able to access this data to get a better understanding of themselves and how it's being used to really create new services for them that they may not even want or something they can overall contribute to in the future? 08:30.160 --> 08:49.840 Well, so you're bringing up a really important point and focus that's near and dear to my heart, which is that as we have all of these new data sources, all of these new capabilities, we also have new challenges as it relates to privacy, a person's rights, 08:49.840 --> 08:53.880 how do we know that information is being used in the way that we intended? 08:53.880 --> 09:00.140 In the context of a clinical trial, I'm an old clinical trial myself, it was quite explicit upfront. 09:00.140 --> 09:01.380 Here's what's going to happen. 09:01.380 --> 09:03.460 Here's how my information is going to be used. 09:03.820 --> 09:09.140 And those controls are already put in place in a very explicit way. 09:10.000 --> 09:19.300 Within the context of the real world data space, when it's around health, we have clear laws as it relates to, for example, HIPAA. 09:19.620 --> 09:39.980 However, as you start to talk about the consumer space, as we start to now think about products, you mentioned the smartphone that has many different sensors and where information may be used in many different ways that are sort of outside of the traditional context of health information for HIPAA. 09:39.980 --> 09:48.060 We now are in a privacy space and a data utilization space that's very different than that which we envisioned 20 years ago, to be honest. 09:48.060 --> 10:10.380 And I think we're at a place where the EULA, the agreements you sort of nonchalantly sign off on as you use different software solutions is probably becoming less and less adequate for making sure that people are fully informed about how their information is going to be used. 10:10.380 --> 10:12.700 So you asked, you know, what's FDA doing? 10:12.700 --> 10:20.620 So FDA doesn't really hold, or under rare circumstances, do we hold identifiable patient health information. 10:20.620 --> 10:26.600 And really, we work under the context of the law, as you can imagine, and that's really not our focus. 10:27.040 --> 10:29.540 So we work in that way. 10:29.540 --> 10:40.780 But we are also, though, and I am very much a part of a social conversation that's really, you know, asking, what does this look like now and into the future? 10:40.780 --> 10:54.980 And I think that as we see these different uses of information, as we also see the different challenges of how it can be misused, it's forcing us as a society to ask what is and is not okay. 10:54.980 --> 11:08.720 Right now, it's largely in the hands of tech companies that we're sort of all these stories are playing out, likely needs to be in the hands of all the different people who really are the core to not only the conversation, but the data. 11:09.800 --> 11:11.280 Exactly, exactly. 11:11.400 --> 11:12.580 That's a really good point. 11:12.580 --> 11:13.720 Thank you for that. 11:13.720 --> 11:28.900 Now, with staying with the tech space, and in terms of cybersecurity, now you have with the advent of the smartphone since 2007, many of the different applications of how we live and interact with life are now onto this little rectangular device that we have in our hands. 11:28.900 --> 11:39.960 And that would also include from a health perspective, insulin pumps, and even glucose monitors, that we can now see the actual overall results of what's happening inside our bodies on our phones. 11:39.960 --> 11:56.860 Now, with many different news articles in the last year, really coming out that these devices could be hacked into from a cybersecurity perspective, what has the FDA done to take safety steps and precautions to ensure that these devices meet the minimum criteria to ensure patient safety, 11:56.860 --> 12:00.940 and this would not happen again in the future, or at least minimize it to some degree? 12:01.320 --> 12:06.200 So as FDA, this is a huge area of concern for us. 12:06.200 --> 12:11.860 And certainly has been a core focus of our devices center, CDRH. 12:11.860 --> 12:18.460 And I commend them, but I also commend the entire community for focusing on this. 12:18.460 --> 12:35.580 And that's been done both through public-private partnerships, as well as specifically through work done at FDA, where we worry about really the development of more secure devices going forward, but also the challenge of legacy devices that still exist. 12:35.580 --> 12:49.440 For example, the insulin pump that was around since many years ago that may now have a security challenge because those have not been fully patched. 12:49.440 --> 12:54.480 And as FDA, one of the things that we've done is build partnerships. 12:54.480 --> 13:01.340 And an example of those partnerships is with actually DEFCON and the biohacking community. 13:01.340 --> 13:22.300 And when I arrived at FDA about 18 months ago, I was pleasantly surprised to learn that not only were we excited attendees of DEFCON, that we also participated in the biohacking community and had been doing so since 2014, I believe. 13:22.300 --> 13:31.800 And that in 2019, we even developed the We Heart Hackers initiative really as a part of our partnership with the white hacking community. 13:32.000 --> 13:51.520 And the interesting thing to me about this work that I've at least learned about is that there have been a number of companies participating, I think over 40 different devices that have been attempted to be hacked into and otherwise patched. 13:51.520 --> 14:00.840 And really has actually led to a community-based effort to start to resolve problems that could otherwise hurt patients if they don't get resolved. 14:00.840 --> 14:06.780 And so to me, this is an example of how the FDA has tried to be forward-leaning into this space. 14:06.780 --> 14:16.980 And instead of just writing guidances and saying thou shall or thou shalt, but actually trying to figure out how we will fix together this particular space. 14:17.700 --> 14:18.500 That's amazing. 14:18.500 --> 14:19.760 That's really good to hear. 14:20.200 --> 14:24.040 That really brings us to the end of our discussion today. 14:24.280 --> 14:37.960 Maybe some last, I guess, parting words for you, Amy, if you want to share with the audience today in terms of what you would tell the hacking community and the attendees of this overall conference and what to do and how best to contribute data so they can, 14:37.960 --> 14:41.920 I guess, better not just devices, but really the overall experience for those that use those devices. 14:42.480 --> 15:02.060 Yeah, you know, I've been thinking a lot about the realities of real-world data, how important I think it's going to be for COVID-19 and how I think that this sets a template for how we may also use real-world data and real-world evidence in the future. 15:02.320 --> 15:09.100 And then that quickly ports me over to the questions around cybersecurity and hacking. 15:09.100 --> 15:11.180 And why do these two come together? 15:11.180 --> 15:34.900 Because if our data sets and our real-world data sets are important for health in the way that helps us understand not only natural history disease, but also what treatments are working or downstream safety events of treatments or vaccines, then those data and the safety of the data themselves and our ability to have confidence that they haven't been tampered with, 15:34.900 --> 15:41.260 that the devices that are generating those data haven't been tampered with, right, is also a critical task. 15:41.260 --> 15:56.280 And so to the white hat hacking community, I ask us to think about how do we continue to keep our real-world data sources safe because those data sources are going to be important for actually our overall public health and safety in the future. 15:56.280 --> 16:02.400 And I think that that's going to be an important area of inquiry and questioning that really hasn't gotten enough attention yet. 16:03.320 --> 16:04.200 Thank you for that. 16:04.220 --> 16:05.240 That's really helpful. 16:05.240 --> 16:13.220 And it's nice to know that this partnership is going to continue and only probably blossom in the future to get even more insights that you can possibly get to, I guess, from now on. 16:13.300 --> 16:18.820 With that, that brings us to the end of this fireside chat session. 16:18.960 --> 16:22.920 So thank you so much for attending and have a great conference, everyone. 16:23.200 --> 16:23.680 Thank you.