Proceedings: DIA Europe 2019

Will New Technologies Accelerate Approval of New Therapies?
Digital Trials and Sensors, Bots and AI, Oh My!

Debra Michaels
Associate Director, Scientific Programs, Americas

DIA Europe 2019

iopharmaceutical development and the broader healthcare industry are faced with growing challenges. The rising cost of healthcare is not sustainable, and the development of biopharmaceutical therapies is too expensive and too slow. Can new technologies such as artificial intelligence (AI), machine learning (ML), “bots,” and other digital tools, help accelerate the pace of development and approval of new therapies? At DIA Europe 2019, a diverse panel looked at the potential impact of new technologies and necessary steps to leverage them.

Impact of Digitization on Acceleration of New Therapies

AI, machine learning, and the like are not new technologies. They have been around for many years, but the biopharmaceutical industry remains about 20-30 years behind other industries in adopting them deeply. We are still operating mainly in an analog world, and just beginning to digitalize our work.

Meanwhile, existing technology is changing at a dramatic pace. It is estimated that the advances made in the next five years will outstrip all those made in the previous twenty! For example, smart phones now have chips with around three billion transistors. A chip with 120 billion transistors–enough to store the structure of all currently known molecules (about 1065)–will soon be available. With our knowledge of genomics, protein structures, transduction of signals in cells, and new AI methods, identifying molecules that can influence the activities of cells can be as rapid as 20 microseconds per molecule. If we can catch up with other industries in applying these powerful technologies, our ability to improve patient lives will be boundless.

Technology as Fundamental Transformation

It is not our technology but our systems, approaches, processes, and attitudes that limit us. To get the benefit of advances in digital technology, we must redefine what an “advanced therapy” means. It must be recognized as a whole package–as a system that includes a medicine and a diagnostic, and mechanisms for self-diagnosis, monitoring, and remote therapy. Conservatism in the healthcare system often arises from regulation and political considerations. We must involve the whole healthcare system in a fundamental transformation from analog to digital, and this will call for careful collaboration among all stakeholders, including patients, healthcare providers, and government.

Access to data is critical, both from an ethical standpoint and in terms of our ability to leverage technology. Without the proper governance, concerns about appropriate data sharing and use will continue to hamper rather than facilitate access. For example, many patients express willingness to share their health data but don’t want it used in a commercial way. We need governance about who should have access to the data and how to ensure fairness and equity for patients and researchers alike.

Further, we need to agree on business models to support further development and use of technology and data. Two types of models are currently operating: a model of central technology development with separate sale of technology and data access, and a model of free technology with monetization and payment for data. Lack of agreement on a universally accepted model contributes to uneasiness around the practice of monetizing data as well as expensive systems that may be less accessible.

True Power of AI: Predictive Applications

Trust in digital systems (in, for example, the algorithms behind AI) is essential for their innovative use in biopharmaceutical development. Biopharmaceutical experts demonstrate one major barrier to building such trust when they believe that they must completely understand the technology in order to validate its output. Other industries, such as the aeronautical industry, employ AI to assist with critical functions such as predicting the useful life of aircraft parts and systems to efficiently maintain the safety of their aircraft. Research and experience have shown these algorithms to be accurate 98 percent of the time, compared to the 91 percent accuracy of the best engineers (methods using Bayesian modeling to employ multiple algorithms in tandem are used to reduce the 2 percent error). The true power of AI is in its predictive applications. It will be necessary to use prognostic algorithms to harness AI to speed up the process of discovery and research (for example, to model the behavior of molecules under various circumstances) and we must develop trust that the algorithms underlying AI will give accurate results.

What is the practical meaning of these ideas, and what are our next steps?

We must find a new and faster regulatory paradigm, away from the sequential regulatory process towards ongoing, continuous learning. This will require a change, for example, in our perspective on benefit-risk: With trust in the information we derive from modeling and simulation methods and other technologies, this new system will provide quicker access to new therapies while continuing to collect real world data that is fed into learning systems to deliver continuous improvement in patient benefit.

The ability to clearly see the outcomes of therapeutic interventions and to positively impact their improvement through this continuous learning system would stimulate creative thinking and allow a more natural healthcare market to evolve.

On a practical level, we must challenge the notion that nothing can be better than the gold standard of the Randomized Controlled Trial (RCT). For example, can we use data from a placebo and standard of care database instead of control arms in every single trial? A TransCelerate project is developing such a database, which now includes data from over 100,000 patients.

We also need methodologies to combine both structured and unstructured data. Instead of trying to structure all data, we can save time and effort by tapping into the value of both. Data sharing will drive process advancement. The technology for building a whole data ecosystem is already here. A data ecosystem that includes open sourcing and data sharing among sponsor, CRO, patient, healthcare provider, and regulators will ensure benefit for all.

Session Chair & Moderator: Isabelle de Zegher, PAREXEL International

Panelists: Julian Isla, Microsoft; Thomas Senderovitz, Danish Medicines Agency; 
Ulo Palm, Allergan; Meni Styliadou, Takeda Pharmaceutical; Peter Shone, PAREXEL Informatics.

Combine Cultures of Technology & Healthcare

We must invest in the development of trust–in technology and in the value of change to build a better system–if we are to take full advantage of advanced technologies in therapeutic development and healthcare. A new way to regulate AI, similar to the way we regulate new drug products, will help to build trust that the technology will work as intended and that its use in the proposed manner is ethical.

We must also combine the cultures of technology and healthcare within educational and training curricula that will prepare our future experts. Each culture is different, and new professionals must have the necessary cross-disciplinary understanding to comfortably blend the two and to create new narratives that encourage multi-stakeholder collaboration.