Artificial Intelligence and Machine Learning in Clinical Trials: Regulatory attempts by European and national authorities
In the last few years, use of Artificial Intelligence (“AI”) and Machine Learning (“ML”) systems in the performance of Clinical Trials (“CTs”) has increased.
Such digital tools can be used to improve the participant selection process, adverse event reporting, therapeutic adherence, feasibility studies, and data quality. In addition, AI technology tools, along with wearable devices (e.g., smartwatches), facilitate real-time, personalized remote data acquisition and monitoring of individual patients.
AI techniques in particular make it possible to analyze large amounts of data. This paves the way for ongoing improvements in the recruitment of participants for clinical trials, so that groups of patients with similar characteristics suitable for a given CT can be identified and those who, for various reasons (e.g., stage of disease, belonging to a specific sub-group), are not suitable for a given CT do not erroneously show up as eligible, only to be removed from consideration later.
In addition, AI/ML tools make important contributions to the monitoring phases of CTs. Raw data from sensors worn by patients can be filtered and processed and then used to make predictions about their clinical condition.
Moreover, advanced data analytics platforms are important. They allow CT feasibility analysis to be performed with a very slim margin for error, as they collect and process very large amounts of information.
Considering this increasingly widespread phenomenon, national and European regulatory authorities are working to introduce a regulation designed to ensure the quality and technical-scientific reliability of data obtained from CTs using these instruments. This is particularly so in light of the purpose of CTs, i.e., to obtain data then used to support applications for marketing authorization for the drugs being studied.
- AIFA Guide
The Italian Medicines Agency (the “AIFA”) recently published on its website a “Guide to the submission of a request for authorisation of a Clinical Trial involving the use of Artificial Intelligence (AI) or Machine Learning (ML) systems” (hereinafter also the “Guide”). This is a useful tool that pharmaceutical companies sponsoring CTs can use to identify the documentation that needs to be submitted with a request for authorization of a CT using AI or ML systems.
The Guide requires that the use of AI/ML systems in a CT be reported at the time of application. In addition, specific information regarding these systems needs to be provided.
A specific benefit-risk assessment of the AI or ML model to be used must be submitted, unless one is already available as part of the research protocol. Additional data to be submitted includes information about (i) the users of the AI/ML model throughout the CT, (ii) the type of model, including the algorithm used and which version, (iii) CE marking, if any, of the device used in the CT.
The Guide provides diagrams that illustrate the application process. In this regard, it should be noted that if the AI/ML model is the subject of clinical investigation within the CT (e.g., for correlation with the objectives and/or endpoints of the CT) the regulatory framework of that AI/ML model in light of relevant Italian and European regulations must be assessed before the application is submitted. In fact, a clinical investigation that concerns the development of a medical device or software intended as a medical device (and therefore an investigation designed to validate that device/software) is subject to the regulations on medical devices and to the supervision of a different authority (not the AIFA but the Ministry of Health).
The AIFA also recommends assessing in advance the necessary documentation for validation, clinical evaluation, and performance of the AI/ML system, and management of information security, datasets, and the format of the data used.
This represents a major regulatory effort on the part of the AIFA within a European framework that has not yet fully embraced digital transformation through appropriate updating of regulations or adoption of specific guidelines on AI/ML in clinical trials (though those are in the pipeline). The AIFA document is constantly updated in relation to new scientific findings and potential regulatory changes.
- EMA Guideline
The European Medicines Agency (the “EMA”) is also considering the regulatory implications of applying digital tools to CTs. On June 10, 2021, it published a “Guideline on computerised systems and electronic data in clinical trials” (the “Guideline”). The document is subject to public consultation through December 17, 2021.
The EMA recognizes the need to provide guidelines to those in the sector and create a specific regulatory framework (there is none in place as of yet) for the use of digital tools, in order to protect the safety of CTs and ensure the reliability of their results.
The Guideline is designed for these specific purposes. It provides useful indications for sponsors, CROs (Contract Research Organizations), investigators, and other parties involved in CTs for the performance of clinical trials using computerized systems. Close attention is paid to the use of wearable devices for the automatic acquisition of biometric measurements (e.g., blood pressure, respiratory measurements, and electrocardiograms), in order to monitor CT participants continuously and in real time and detect clinically relevant data.
The document outlines the basic principles to be applied to all computerized systems used in a CT. These include the principle of data integrity, proper data storage, and clear and specific identification of the roles and responsibilities of the parties involved in the CT. The sponsor also must guarantee adequate qualification and validation of the electronic data processing systems used in the CT.
- What are the next steps in regulation?
The evolution of digital tools and their fundamental contribution to CTs require specific regulation at both a national and European level. With this in mind and with the aim of simplifying the development of innovative technological solutions and making them available to patients in a safe way, specific expert groups, such as the Innovation Task Force (the “ITF”) and the Scientific Advice Working Party (the “SAWP”), have been set up at the European level. The ITF is a multidisciplinary group of members from the EMA and national regulatory authorities engaging in early dialogue on innovative technologies, while the SAWP provides specific advice for innovative drug development tools (such as digital technologies).
In addition, as mentioned above, the EMA focused on providing a specific guidance document for the use of AI in CTs in the form of the Guideline. This document provides a reference framework, and the various national initiatives on this matter should be included and coordinated within it. With application of the new Regulation (EU) no. 536/2014 on clinical trials approaching, a coordinated and uniform method at the European level is needed more than ever.
 Please refer to the following paper: https://www.osservatorioterapieavanzate.it/innovazioni-tecnologiche/digital-health/l-intelligenza-artificiale-al-servizio-dei-trial-clinici.
 For more information, visit https://crasecrets.com/tecniche-di-intelligenza-artificiale-applicate-alla-ricerca-clinica/.
 For in-depth analysis, see https://crasecrets.com/tecniche-di-intelligenza-artificiale-applicate-alla-ricerca-clinica/.
 The Guide is available in both Italian and English at https://www.aifa.gov.it/-/guida-alla-presentazione-della-domanda-di-autorizzazione-alla-sperimentazione-clinica-che-preveda-l-utilizzo-di-sistemi-di-intelligenza-artificiale-ai-o-di-machine-learning-ml-.
 Regarding the application of EU Reg. 2017/745 to clinical investigations related to MDs as of May 26, 2021, please see the Ministry of Health Circular of May 25, 2021.
 For more information, see the AIFA “Guide to the submission of a request for authorisation of a Clinical Trial involving the use of Artificial Intelligence (AI) or Machine Learning (ML) systems.” Link in footnote 4.