AI: Reconciling the Need for Standardization vs. Customization
In the last edition of #KCR_Trends, we discussed how Artificial Intelligence (AI) became the newest hot topic in clinical research. The industry has noticed and is currently heavily investing resources into AI to leverage its potential for use in clinical trials. AI is a broad term that encompasses a variety of different things: Data Mining, Neural Networks, Machine Learning, etc. Each of these tools works by utilizing “structured data” (i.e. information organized in a rigid way to ensure adequate analysis as well as trend & pattern recognition). The article will differentiate between advanced analytical capabilities and AI. The industry has decided that AI sounds much better, therefore we will keep it that way.
So how, in the modern world of pharma, with increased complexity in study designs and the pursuit of ever-more rare indications, can tools relying so heavily on standardized information work in a world of “unstructured data” and customization?
Most analysts agree that currently AI can support clinical trial execution by doing 3 things:
• Cutting processing time
• Informing decision making
• Improving data analysis
Handling processes efficiently
AI is a broad term that encompasses a variety of different things: Data Mining, Neural Networks, Machine Learning, etc.
A JP Morgan analysis estimates that approximately 80% of trials will suffer recruitment problems at some point during the study. To combat this, companies pour millions of dollars into recruitment strategies to see little payoff. But now there are new solutions: half a dozen firms are already working with AI applications to identify and match patients to trials in real time. According to Kumba Sennaar, TechEmergence, some have already considerably cut recruitment times and have exponentially improved patient engagement within their trials by matching patients to studies relevant and close to them.
Protocol design through AI
As mentioned above, R&D industry continues to focus on rare and orphan diseases with an associated increase in protocol complexity. In this context, AI can analyze large data sets with information from previous trials, spot similarities and patterns, and inform protocol design to maximize the chances of successful development programs. Some industry experts believe this predictive data can also help companies detect which patients have higher chances of dropping out of the trial and who will be more or less likely to adhere to the treatment plan. A case study by one provider, for instance, claims to have reduced timelines by a third.
Trial intelligence & speed
The pursuit of more advanced therapies, such as biologic treatments, often increase the number and complexity of endpoints to ensure clinical efficacy, so AI will likely increase regulatory oversight. Data points in clinical trials have become increasingly more complex in recent years, whereas 30 years ago trials were conducted entirely on paper. Not only is this not environmentally conscious, but it would also be impossible to manage the sheer amount of data in a non-digital way. For instance, some sponsors conducting trials on Alzheimer’s are including voice recordings as endpoints to analyze speech differences in patients and progression of the disease.
Several hurdles remain in the way of AI to be used in clinical trial execution, the most relevant one being the regulatory environment. Yet, authorities have begun to understand the potential for patients to benefit from AI tools and solutions, and have taken steps to ensure an adequate environment exists for AI to grow.
