Postmarketing requirements (PMRs) for drug development are commonly required to gather data on a product’s longer-term safety, efficacy and optimal use. The probability of postmarketing requirements depends on a variety of factors; no category is insulated from them and requirements are significant across the board. Drug developers crossing the US and European markets can face differing PMRs and timelines from the FDA and EMA, adding compliance complexity along with additional costs. Given drug development costs averaging from under $1 billion to $2 billion, unexpected costs or delays can make or break a new drug’s commercial success. Carefully applied AI and machine learning offers the potential for better management of postmarketing requirements.

AI is already transforming drug development

AI is already showing a positive impact on many areas of drug development, from trial site selection to predicting a potential drug compound’s efficacy. One of the major benefits of AI and machine learning (ML) is moving a sponsor from being trapped in information overload to making better informed decisions.

AI and ML can better process the massive amounts of data related to multiple aspects of drug development faster and more efficiently incorporate new data faster than would otherwise be possible. From this information, the technology can identify patterns which generate valuable and actionable insights.

Combined with expert human insight, the technology provides a never-before reached level of sophistication and capability. In the case of PMRs, AI tools have the potential to predict the likelihood of additional requirements with extreme accuracy and the types of assessments likely to be needed.

Clinical trial designs that start with the end in mind

The ability to accurately forecast a PMR means that developers can better plan budgets, resource use and timelines for those eventualities. One benefit of early identification includes the ability to plan for and gather supplementary information earlier.

Clinical trial tokenisation is a method of de-identifying personally identifiable information (PII) and easing the burden of postmarketing requirements. An encrypted token replaces consenting patient identifying details. The patient token can then be connected to real world data from other systems and serve as a reference for a wider and longer-term picture of relevant health topics. Sponsors can identify trends to help address postmarketing requirements. Early implementation avoids later delays and significant unexpected costs. Clinical trial tokenisation complies with good clinical practices and Health Insurance Portability and Accountability Act (HIPAA) guidelines.

Cassandra: A highly accurate AI tool to predict PMRs

ICON’s AI solution, Cassandra, uses data from drug applications, approvals and rejections drawn from FDA, EMA and Citeline dating back to 2003. As of March 2024, Cassandra contains over 220,000 individual drug records and nearly 435,000 trial-related records covering 261 therapeutic classes and 3,912 mechanisms of action for more than 103,000 primary drugs globally. Cassandra’s evaluations incorporate more than 3 million data points, providing an expansive and accurate experience base. Information on existing and new entries is updated quarterly as new data is provided by the FDA and EMA commitment databases.

Using proprietary algorithms, Cassandra evaluates the likelihood that a drug will necessitate postmarketing requirements and the type of information that regulatory authorities will seek. It does this by considering the same and similar molecules and mechanisms alongside previous regulatory actions.

Artificial intelligence and machine learning offer the most value when leveraged with human insights and oversight. Results from Cassandra are always validated by ICON experts in real world solutions, scientific affairs, therapeutics and drug development services. Cassandra has accurately predicted 99% of FDA PMRs and 97% of EMA PMRs.

Conclusion

When paired with human expertise, AI tools can help provide crucial insights into postmarketing requirements earlier in the development cycle. The accuracy of those insights is reliant on the data quality and quantity, as well as the processes used to manage and process the data. These insights can be used to better manage development, including mitigating the risk of additional expenses, delayed commercialisation and the loss of a timing advantage. By applying AI with skill, they can set their trials on a course to success from the outset.