Impact of AI on Clinical Development

December 10, 2019

In association with Statisticians in the Pharmaceutical Industry (PSI) , UCB and Cytel hosted a symposium on September 11, 2019 at UCB’s offices in Slough, Berkshire. The primary agenda was to educate the audience on Artificial Intelligence (AI) approaches and their impact on clinical development.

With recent advances in AI, it is important for quantitative scientists to keep up to date with the most recent methods and be involved in guiding their application to the most pressing analytical challenges. This one-day event covered cutting edge examples of how data science and statistical sciences are intersecting, and its relevance to our attendees.

“Artificial Intelligence and associated methodology is becoming increasingly important to the Pharma Industry and its technical foundation in statistical theory means that PSI is naturally keen to promote good practice through its membership and established Industry links. PSI is proud to have set up a Special Interest Group in this field and is keen to broaden its links and membership.”

- PSI Data Science special interest group

In this blog, we share some of the key takeaways from the symposium. If you are interested in attending similar sessions, you can check Cytel’s list of upcoming events here.

Dr Karim Malki, Head, Predictive Analytics, UCBDr Karim Malki, Head, Predictive Analytics, UCB

Involving statisticians in AI and machine learning (ML)

Moira Verbelen Principal Statistician, Predictive Analytics at UCB spoke about the critical role of statisticians in AI and ML projects.

Statisticians have an essential role to play in the AI and ML projects planned by their organizations. In particular, they can ensure that the correct methodology is used, which also involves gauging the need to use AI and ML in a study. Statisticians in collaboration with other multidisciplinary teams can contribute significantly towards realistic interpretation of results. Statisticians are well placed to assess what a result means and, importantly what it does not mean. This will ultimately save time and money for an organization and ensures that the data science team invests its time in well-designed projects.

To effectively create AI and ML analyses, statisticians need to begin by understanding data structures and the model designs. In fact, basic ML models are very similar to traditional statistical models. For example, LASSO and elastic net are ML extensions of linear regression.

It is an exciting time to be a statistician and play this essential role in forging the new era of data science. As statisticians develop new data science skills, they will be able to participate in the growing number of AI and ML projects within organizations.

 

Developing Advanced Analytics Communities

Chris Harbron an Expert Statistical Scientist at Roche spoke about Roche’s experience in developing Advanced Analytics Communities both Internally and Externally.

A large volume of healthcare data is available to us today through a variety of sources, in real-world and clinical trial settings, that needs advanced analytics to unlock its potential. However, organizations end up investing significant time in expert curation and preparation of the data before advanced analytics can be applied.

Roche is building its personalized healthcare capability by developing internal and external advanced analytics communities. Aimed at finding better treatments for patients through systematic knowledge exchange, data scientists at Roche established the Roche Advanced Analytics Network (RAAN) in 2017. It is a global community with over 900 members across more than 40 Roche sites. For the external environment, Roche developed an Advanced Analytics Academic Outreach Model to initiate collaborations with leading academics and academic institutions focused around methodology development. With this, the company seeks to be at the forefront of new AA methodologies and approaches.

Strategic collaboration is at the heart of creating empowered advanced analytics communities. Access to meaningful data coupled with expertise in applying fit-for-purpose analytical approaches is key to enabling new discoveries. By encouraging knowledge sharing and collaborating to develop your AA expertise to create insights from data, we will ultimately improve patient care.

 

Promising health applications of AI and ML and their challenges

Dr Chris Holmes from the University of OxfordDr Chris Holmes from the University of Oxford spoke about a few promising health application areas and the challenges of using AI and ML in healthcare.

AI is set to transform medicine as it is being successfully applied to healthcare challenges. We are witnessing advances in digital measurement technologies such as new applications for image analysis to detect and diagnose diseases. Developing natural language processing techniques to analyse electronic health records has created a better understanding of drug efficacy. AI tools are also being efficiently used at healthcare facilities for activities like scheduling appointments, checking systems and managing wait times. Wearables and consumer devices are being commonly used today to manage personal wellbeing.

Using AI and ML in health comes with its own set of challenges. In healthcare, data generation and data capture are no longer the bottlenecks. However, this data is often heterogeneous and is collected from multiple sources- these are challenges that need to be overcome. Some of the AI and ML models are quite complex and are largely black-box models for users. This impairs their reproducibility and explainability. That, in turn, hinders their adoption in highly regulated industries.

Despite the challenges, AI and ML seem to be the hottest technologies in healthcare. Machine learning methods, such as deep learning are incredibly powerful and useful tools. But they need to be seen as support tools and not as substitutes for careful thought.

Francis Kendall from CytelFrancis Kendall from Cytel was also present at the event, his talk focused on the paradigm shifting approaches we are witnessing as new players enter clinical development. As large tech companies gather more and more health data, they are able to approach the challenges to drug discovery, monitoring of trials, and engagement with clinicians in ways that were previously impossible. With these changes, traditional pharmaceutical companies need to reposition themselves in order to both engage with these new tools and develop their own access to data to maintain competitiveness.

 


 

Rob Power, Consultant Statistician, UCB and Panel

Panel (from left to right): Chris Harbron, Expert Statistical Scientist, Roche, Dr Moira Verbelen, Principal Statistician, UCB, Francis Kendall, Senior Director, Biostatistics and Programming, Cytel, Rob Power, Consultant Statistician, UCB

“Artificial Intelligence and Machine Learning are playing a vital role in transforming the healthcare landscape. At UCB, we are continuously working to advance science and embrace new technologies. We’re leveraging cutting-edge techniques to deliver value to patients when it comes to their care. UCB believes strongly in partnership and the AI symposium was a great platform to collaborate with the industry experts.”  - Rob Power, Consultant Statistician, UCB


 

Download this ebook to see examples of how Cytel's data science and real-world evidence groups have helped clients apply advanced analytical techniques to large, complex historical or real-world data sets to improve their decision-making, accelerate development pathways, and enhance their probability of success.

download ebook