Could you provide an overview of how AI is being utilized in the development of new cough medicines, and what specific challenges or opportunities it addresses in this field?
AI is transforming many fields and preclinical drug discovery, design and development are no exceptions. However, acoustic AI is playing a particularly important role in the clinical development of new drugs. The crazy thing about clinical drug development is that, in studies done to date, cough as a symptom is rarely quantified and even when measured, it is done so inaccurately.
Thus, the simple ability to continually and unobtrusively monitor coughing is a game changer in development of drugs that have cough reduction as a primary endpoint or cough induction as an adverse event. Recent advancements in cough science are yielding insights into the natural history of cough that helps with study design and into the statistical patterns of cough that are critical to demonstrating how those change with therapy.
What novel AI technologies or methodologies are being employed in the research and development process of cough medicines, and how do they enhance traditional drug discovery approaches?
The current approach to cough monitoring only provides the number of coughs over 24 hours pre- and post-dosing. However, new technology has been developed by applying deep learning models to massive cough data sets to create algorithms that recognize and time stamp coughs. These algorithms run in near real-time entirely on devices enabling privacy-preserving continuous cough monitoring. The resulting series of timestamped coughs is extremely rich data for looking at baseline cough patterns. The application of both standard and proprietary statistical approaches to these data provides a very sensitive approach to detecting drug effects.
What are some notable benefits and drawbacks of integrating AI into the development of cough medicines, particularly in terms of efficiency, cost-effectiveness, and efficacy of the resulting treatments?
In November of 2023, Merck presented efficacy data on a novel P2X3 inhibitor to CDER in the form of 24-hour cough counts at several points in the trial and the FDA denied its approval. The transcripts of that meeting provide some key insights into the limitations of such an approach. These included a lack of data on inter-observer variability, concerns with using evolving data compression algorithms, the inability to detect or report on bouts as a measure of cough severity, and the relationship between PROs and objective cough monitoring.
Interestingly, the biggest problem with the way cough counting has been done - cough variability - was not mentioned in the advisory committee meeting.. There are now data supporting what every cougher knows - these folks have good and bad cough days, but Pharma is rolling the dice by measuring only 24-hour periods before and after dosing. If they happen to enroll a subject on a low cough day, they miss the therapeutic impact. And if they enroll the subject on a high cough day, they will overreport the therapeutic effect. Furthermore, there was a huge placebo effect, some of which we believe will be reduced with more accurate endpoint assessment.
Given the regulatory landscape surrounding pharmaceuticals, what are some key considerations or challenges when it comes to incorporating AI-driven methodologies into the development and approval process of cough medicines?
The FDA has not approved a new antitussive in over 60 years and it seems during that time the agency and cough experts have lived in parallel universes. From population biology we know what happens under these circumstances: intellectual speciation. I have great respect for both groups, but they need to get better aligned if we are to see new drugs through the regulatory pipeline.
Being new and science-based, I believe that the quantitative and statistical rigor of continuous cough monitoring can accelerate that alignment and I would be thrilled to see FNIH, DIME, the Critical Path Institute or others take the lead in brokering that conversation.
How do you envision the future of AI-driven innovation in the field of cough medicine development? Are there specific areas or applications within this realm that hold particular promise or warrant further exploration?
There has been a lot of attention to cough suppression, but cough is a problematic symptom that troubles patients afflicted with over a dozen conditions - asthma, COPD, lung cancer, CHF, and IPF, to name only a few. To date we have been caught in a vicious cycle - it could not be measured so providers and drug developers paid it no attention. This is rapidly becoming a virtuous cycle with investigator-initiated studies to better understand its causes and management and industry designing better medical devices and molecules to treat it.
I’m particularly excited about a digital therapeutic for cough. It is a poorly kept secret that spending four hours with a trained speech pathologist can substantially reduce cough in 70% of people with UCC/RCC and we are currently digitizing that curriculum to be delivered through a phone app and rigorously testing to verify that it works, and therefore greatly expanding the access to this toxicity- free intervention.
In what ways does the use of AI in cough medicine development contribute to personalized or precision medicine approaches, and how might this impact patient outcomes and healthcare practices?
We have just begun to scratch the surface of cough science. For example, it’s been proven that there is an enormous discrepancy between objective cough counting and patient-reported outcomes. At one level that is unsurprising - the former is quantitative - and an excellent endpoint for pharmacodynamic studies, whereas the latter addresses how cough makes one feel. Rather than arguing which is “better,” I’m interested in what we can learn from these orthogonal views. I believe we can use them to understand what it is about cough that is bothersome to an individual patient. For one, it may be the resulting urinary incontinence; for another, it may be social embarrassment. Knowing this and targeting therapy accordingly will move alleviating the suffering of coughing into the age of precision health.
Can you discuss any ongoing or upcoming research initiatives or collaborations aimed at leveraging AI to address unmet needs in the treatment of cough-related conditions, and what potential breakthroughs or advancements they may yield?
We are in the golden era of cough. COVID-19 elevated the general public’s awareness of cough and further increased the stigma of those of us who suffer from it. Fortunately, there is a new understanding that cough is not always a symptom - it can be the disease itself, cough hypersensitivity syndrome, a condition as common as asthma. Scientists have identified the neural pathways of cough and Pharma is developing selective inhibitors that hold great promise to become the first truly effective new antitussives. Finally, the availability of affordable and fieldable cough monitors have spawned an explosion in cough-related clinical investigations addressing a very broad range of conditions. On a personal note, I am excited to see that much of this work is being funded by organizations such as the Gates Foundation to address public health threats such as tuberculosis in under-resourced countries.
Author Details
Dr. Peter Small, M.D. Chief Medical Officer - Hyfe AI
Publication Details
This article appeared in American Pharmaceutical Review - Innovations at Interphex 2024 Supplement
Pages: 24-25
Subscribe to our e-Newsletters
Stay up to date with the latest news, articles, and events. Plus, get special
offers from American Pharmaceutical Review delivered to your inbox!
Sign up now!