Artificial Intelligence in Clinical Studies

Clinical Trials

Artificial Intelligence (AI) is a growing field of interest for many industries, including those of medical devices and pharmaceuticals. Clinical studies have an increase in focus on patient-reported outcomes which can provide large datasets. AI could prove to be a valuable tool in streamlining clinical research.

Potential applications of AI in clinical studies

Automating processes and data analytics can enable consistency and efficiency throughout a study, while contributing to patient safety.

Slow enrollment is a common problem in many clinical studies. Incorporating AI tools such as natural language processing during screening can aid in registry database and medical record searching. Many recruitment campaigns on social media already use AI and machine learning algorithms to target a specific patient population through demographic information and search history.

With the growing popularity of electronic data capture systems (EDC), data safety monitoring can be performed in real time through automated reports. For example, the EDC can be designed such that the sponsor will receive an email if a serious adverse event is entered. AI is also being used for predictive modeling for estimating the probability of adverse event occurrence and treatment effectiveness. To decrease time to submit data, AI can be used to de-identify images and medical records. This can help the study team to more quickly respond to patient safety concerns.

Digital health technologies often use AI for the collection and processing of patient physiological data. More passive data collection can help improve retention by reducing patient burden and can also offer the issuing of automatic safety alerts to the site and reminders to the patient.

Overcoming challenges of AI

Interoperability is a critical aspect of AI integration. The use of controlled terminology aids in being able to process data across different formats. For example, the use of terms based on Clinical Data Interchange Standards Consortium (CDISC) standards can be used to organize data. Eligibility criteria and adverse events can be coded using Common Terminology Criteria for Adverse Events (CTCAE) or Medical Dictionary for Regulatory Activities (MedDRA).

Machine learning uses AI to develop models based on training algorithms and datasets that are not explicitly programmed, which can make verification and validation difficult. The American Society of Mechanical Engineers (ASME) developed standard V&V 40 to assess the credibility of computational models for medical devices. Clinical studies of software as medical device or studies that use AI algorithms for endpoint assessment may include elements of external verification and validation. For example, a study of an investigational wearable device that uses AI to determine the probability of getting sick might involve the patient using a thermometer, oxygen finger sensor, or Holter monitor, along with a daily diary of symptoms to verify that the collected parameters from the wearable are accurate and correlate with the patient’s symptoms.

Regulatory oversight

In the US, there is currently no regulatory framework for the use of AI in clinical studies. However, the FDA established the Digital Health Center of Excellence and publishes discussion papers and podcasts about the role of AI in clinical research.

The EU has endorsed the Artificial Intelligence Act, which includes AI systems used in medical devices. The EU Medical Device Regulation requirements align with the EU AI Act. These include overarching principles that can be applied to clinical studies.

Conclusion

Understanding the capabilities and limitations of AI tools can aid in efficient clinical study conduct and management. As the use of AI in clinical studies grows, attention must be focused on complying with new regulations and best practices.

MED offers a variety of clinical trial services and is a full-service CRO. We have over 40 years of experience designing and executing clinical trials, ranging from early feasibility studies to multinational, controlled pivotal trials to post-market registries.

Contact us to discuss how you can integrate AI in your clinical study!