Increasing Sophistication Requires Strategies for Adoption
As clinical trials have become more complex with more moving parts, so too have the requirements for technology specifically automation. Automation comes in various forms; new capabilities, integrations, and Robotic Process Automation (RPA) are examples. These technologies provide a consistent, deterministic means of automating repetitive and predictable tasks. As automation needs become more complex, the sophistication required of automation has also increased. Artificial Intelligence (AI) and Machine Learning (ML) have aided in managing this complexity through “learning while automating”.
RPA enables us to quickly automate processes to the extent that the initial automation can be iterated upon, continually improving and extending the positive impact. For example, at ICON we have enabled automation using RPA across a wide range of disciplines: finance, Trial Master File loading, locking of clinical databases, and managing help desk tickets, to name a few. These processes are both clinical and administrative in nature, enabling consistency, timeliness and quality to remain as the focus.
A key challenge to the adoption of automation in clinical trials is the preservation of compliance with regulatory requirements, particularly as RPA can be continually updated as business needs change and new requirements are discovered.
Some key components of the approach described by the FDA in their Proposed Regulatory Framework for Modifications to Artificial Intelligence/Machine Learning document, published in 2018, are:
1. The establishment of a quality system and best practices for development.
2. Demonstration of safety and effectiveness through a pre-market review so that patient risk can be managed throughout the lifecycle.
3. Ongoing monitoring of the device to incorporate risk management in the development, validation, and execution of any changes, managed through tightly controlled change management practices.
4. Use of post-market real-world performance reporting to maintain the continued assurance of safety and effectiveness.
The discussion document from the FDA focuses on Software as a Medical Device. The same concepts could be applied to a validated process for the implementation of sophisticated RPA automation. Given the speed of development associated with RPA, how can it be controlled, and the evidence gathered in such a way to prove the efficacy/benefit associated with the specific automation? Similar decision points could be used to determine what level of re-validation may need to be executed based on updates to inputs, processes/algorithms and intended use.