Though it’s not often discussed, often the most difficult aspect of bringing a drug to market is finding the right patient population. This is especially true with rare diseases, where patients are routinely misdiagnosed or undiagnosed. With very low incidence rates and a population randomly spread across the United States, finding patients to can be daunting, if not nearly impossible.
Audentes, an Astellas Company, was facing this challenge. The company wanted to bring to market their gene therapy for the treatment of X-linked myotubular myopathy, or XLMTM, but were burdened with the inability to find patients. XLMTM affects newborn males exclusively and has a 50% mortality rate within the first 18 months of life and there is currently no available treatment for this devastating condition. The statistics paint a pretty clear picture of how rare this disease is – only 40 boys were afflicted with XLMTM out of the approximately 2 million males who were born last year, though there may be patients who are yet to be diagnosed. To further complicate finding, tracking, and treating patients, XLMTM has no ICD-10 code.
To tackle this problem and move forward with their critically-needed drug candidate, Audentes enlisted IPM.ai. What happened when these businesses came together was explored via our webinar hosted by Jeremy Smith, Director of Business Analytics for Audentes along with Dan Fisher, Principal Consultant at IPM.ai.
Dan explained how we leveraged machine learning and artificial intelligence to create an algorithm that could identify XLMTM patients. To make this possible, we relied on our extensive data universe of over 300 million de-identified unique patients, complete with over a decade’s worth of their medical history and demographic markers (including 3,700+ consumer attributes and 65 billion consumer records). However, given the very young age of patients, perhaps the most important factor of our data was not SDOH but its near real-time availability – our data streams refreshed weekly and are updated with every new medical event, translating to unparalleled real world insights.
We began patient sourcing by importing and scoring 76 confirmed positive patients, given to us by a leading XLMTM researcher’s registry. Our algorithm was engineered to recognize key events of the condition and work backward from eventual diagnosis — starting with the first instance of respiratory distress through hospitalization. After analyzing billions of interactions and scoring our entire patient population, we were able to identify ~ 10% of all XLMTM cases as well as the health care practitioners treating them. Attendees of the webinar had a variety of questions on this subject, ranging from what we did next to informed consent. Dan Fisher responded by noting that the next phase of the project includes finding and alerting Audentes about incremental patients who have been coded with the new XLMTM ICD10 code. As for consent, it was not needed for IRB approval likely due to patient age. However, each step in our process is designed to maintain compliance with all relevant regulatory bodies and in collaboration with internal legal and compliance teams.
Follow-up questions included validating token sample size (though seemingly low, 76 is actually a robust figure given the overall incidence rate) and targeting PCPs. “We typically provide our clients with the most recent specialist and most frequent provider…understanding which pediatrician or family medicine doctor is managing the overall case for the patient can influence referrals and treatment initiation. We spotlight these providers at the NPI level, which can be fed into personal outreach (sales, MSL) or non-personal channels such as targeted HCP digital media or email campaigns,” Dan explained, adding that this is a crucial yet often missed step.
Patient population is often taken for granted, but finding the right patients, especially for emerging therapies and in rare disease is pivotal for getting life-saving drugs to those who need them most. Our work with Audentes demonstrates the necessity of RWD in identifying, diagnosing, and treating patients; quite simply, actionable data has become key to moving a drug forward.