The recent 2018 Clinical Trial Innovation Summit had a session titled: “What Are the Opportunities and Challenges with Applying Machine Learning and AI to Clinical Trials?”
A panel of experts from Bayer, Roche and Janssen discussed how AI and Machine Learning can be employed in clinical trials. While there were many naysayers in the audience with individuals fixated on the negatives and the “fuzziness” of AI and Machine Learning, I felt different – more positive about the topic. In a past life as CTO of Thomson Reuters Scientific, we implemented a very effective classification tool that did a good job of mimicking human coders. It was a classification project and not regulated in any way. If the code was wrong, we would get berated by the customer, acknowledge the error and fix it. We are talking about a completely different situation here with very serious consequences, so a heavy dose of caution is warranted.
“As long as there exists a computer that never tires and is capable of looking for patterns in clinical data at the speed of light and can short list the data in need of review by the clinician, there will be a need and value associated with AI to expedite studies.”
Based on this and other past experience, I can say that AI, not defined as Artificial Intelligence that implies “replacing” humans but rather defined as Assisted Intelligence, definitely has room in clinical trials. As long as there exists a computer that never tires and is capable of looking for patterns in clinical data at the speed of light and can short list the data in need of review by the clinician, there will be a need and value associated with AI to expedite studies. I can already hear those who have done research in this field express concerns about false positives and false negatives which is often the Achilles Heel of all such AI/ Machine Learning Systems. I am in agreement. However, it’s ALL about the quality of data that is fed into these systems.
Francis Kendall, a speaker from the AI session, suggested that our biggest problem is data quality. He is absolutely correct. Over the years and through the EDC revolution, we have adopted many point solutions to collect data. Unfortunately this has resulted in a massive fragmentation of data. These AI/ Machine Learning Solutions are busy understanding the data and looking for patterns, and the last thing that they need is data that is inconsistent and not clean. This is exacerbated by the fact that training data is often pristine and carefully collected and curated to facilitate the “learning” part of Machine Learning. So it is no wonder that when poor quality or inconsistent data is fed into such mechanisms, the results end up being just as poor. Thankfully there are solutions that can apply intelligent transformations, taking into account the context of the data at the time of capture through utilization of a smart middleware that pulls from data sources and conforms to a format that is easily digestible by all such systems. Kendall concluded that as we achieve data conformity, we will grow to accept AI and Machine Learning usage in Clinical Trials…I totally agree with that as long as we treat AI as Assisted Intelligence and not relinquish all faculties to it!
Sina Adibi is the CEO of Adaptive Clinical Systems and a veteran of Pharma and Health IT industry.
About Adaptive Clinical Systems
Adaptive Clinical Systems offers a unique, simple, secure, validated, compliant, and cost-effective innovative solution for clinical data integration and interoperability. The cloud-based innovative Adaptive eClinical Bus® solution integrates clinical study data from multiple systems and platforms — EDC, eCOA, CTMS, Medical Imaging, IRT, analytical/data visualization systems and others — to ensure accurate and efficient transfer of clinical data for any study of any complexity while going well beyond simple and difficult to scale integration to full, real-time interoperability.
The award-winning Adaptive eClinical Bus software includes “connectors” for many leading clinical trial software tools from well-known vendors such as Omnicomm, Medidata, BioClinica, and Clinical Conductor to open source clinical trial tools such as OpenClinica and Clinovo. Connectors can also leverage internally-developed and proprietary systems and help customers retain their competitive edge. Adaptive Clinical’s eClinical Bus® can easily integrate technology into an interoperable, efficient, and accurate clinical trials system that streamlines processes and improves data reliability and offers the freedom to choose the best eClinical tools of any third-party or proprietary systems while enjoying all the benefits of a fully integrated system.