Improving Clinical Trial Productivity using AI/ML

Context

Clinical trial is an essential and important step in the drug development process to demonstrate a potential drug’s safety and efficacy to gain Food and Drug Administration (FDA) approval. FDA typically requires Phase 1, 2, and 3 trials to be conducted to determine if the drug can be approved for use. The Pharma industry, on average, spends ~2B USD to bring a new drug to market, of which a large part of the spend is on clinical trials, especially the phase 3 trials. And considering that not all potential drugs that undergo clinical trials gain FDA approval, it is important for Pharma companies to explore opportunities to improve Research & Development (R&D) productivity and reduce cost.

Apart from the regulatory needs, there are a lot of system inefficiencies and manual processes in clinical trials that add to the cost and time to bring new drugs to market. Over the years, Pharma companies have adopted various digital technologies such as Cloud, Mobile apps, Robotic Process Automation (RPA), Big Data, Advanced Analytics, Artificial Intelligence (AI) and Machine learning (ML) to automate clinical trial processes, shorten timelines and reduce costs. The emergence of Generative AI (GenAI) has further opened new opportunities to automate and transform clinical trial processes and thereby improve R&D productivity and ROI.

AI/ML Use Cases

Significant progress has been made over the years in digitizing and automating the clinical trials process with the adoption of systems such as electronic data capture (EDC), electronic informed consent (eConsent), and electronic clinical outcome assessment (eCOA). But there are still many inefficiencies and several manual processes such as protocol design and document creation, study setup, data management, and clinical study report (CSR) creation.

AI/MLpowered capabilities such as pattern recognition, natural language processing (NLP), named entity recognition (NER) and extraction, predictive analytics and now GenAI-based content creation can help resolve some off these issues and in-efficiencies. Let us look at some possible use cases for using AI/ML in Clinical Trials.

• Trial Design:  All Pharma companies have a lot of historical data regarding the clinical trials conducted in the past. In addition, trial data is also publicly available in sites such as ClinicalTrials.gov. AI/ML can be applied to learn from past data and design optimal future studies/trials, thereby increasing the chances of completing the trials successfully.
• Protocol Creation and Digitization: AI-based protocol builder can create the protocol document based on the information from the past trials and using the standard protocol template developed by Transcelerate. In case of ongoing trials, the existing protocol documents can be digitized, and relevant information such as inclusion/exclusion criteria and endpoints/objective can be extracted using AI/ML and made available to downstream systems for further processing.
• Site Identification and Selection: Identifying and selecting the right sites for the trial is a critical trial startup activity. Apart from the experience and qualification of staff, availability of specialized diagnostic equipment, Geographic location, and past track record of sites with similar previous trials is an important criterion for site selection. AI/ML can help analyze the historical site data and identify the best sites for conducting the trial.
• Patient Recruitment: Around 80% of global clinical trials fail to identify and recruit eligible patients on time. This has a direct impact on the time to complete the trial as well as additional financial burden. AI/ML can be used to analyze patient population data from public databases and social media content to identify specific regions/countries where a particular disease/condition is more prevalent, and help define the recruitment strategy. AI/ML can also be used to identify the eligible patients from EMR/EHR records based on the selection and inclusion exclusion criteria specified in the protocol document, thereby accelerating the recruitment process.  
• Patient Retention: Several patients (~30%) dropout and do not complete the trials due to various reasons such as large number of visits, inconvenience, lack of support and others. AI/ML-based assessment can be performed to identify participants at high risk of protocol non-compliance or drop-out at the start and during the trial. Such participants can then be provided with the required intervention, support and engagement, thereby improving patient experience, protocol compliance and retention.
• Clinical Data Management: While conducting the trial, data is collected from different sources/systems, validated, cleaned, and finally converted to Clinical Data Interchange Standards Consortium Study Data Tabulation Model (CDISC SDTM) standard. AI/ML can be applied to streamline clinical data management by automating data collection, validating and monitoring data quality, raising, and resolving site queries. In addition, the SDTM conversion process can also be digitized with AI/ML-based algorithms performing an automated mapping from the source dataset to SDTM.
• Patient and Site Assistants: AI-based conversational assistants can be deployed to assist, facilitate, and engage both Patients and Site staff during the clinical trials.
• Trial Master File (TMF): A lot of documents are created during the clinical trials that need to be stored and retained for regulatory purposes. OCR and AI/ML-based technology can be used to digitize and automatically store the trial documents in the correct eTMF folder.
• Pharmacovigilance (PV): Pharmacovigilance is a set of activities related to the detection, assessment, understanding, and prevention of adverse events. AI/ML can applied to automate the process of safety case creation, case evaluation, including causality assessment, and individual case safety report (ICSR) creation.
• Clinical Study Report (CSR) creation: Clinical study report creation is a manual and highly time-consuming process. The contents of CSR are derived from the protocol, statistical analysis plan, safety narratives and other sources. GenAI-based solution can be developed to automatically create CSR documents by pulling data from various clinical systems and records.

Regulatory Implications

As with any innovation, while AI/ML creates new opportunities, it leads to new and unique challenges. To address these concerns, the FDA has released a discussion paper, “Using Artificial Intelligence and Machine Learning in the Development of Drug and Biological Products.”[2] This paper aims to engage with industry stakeholders to solicit feedback for developing new regulations and guidelines for the use of AI/ML in drug development while safeguarding public health.

In this paper, FDA has provided an overview of the current and potential future uses of AI/ML in drug development process. It also discusses the possible concerns and risks associated with the use of AI/ML, specifically in areas such as explainability, reliability, privacy, safety, security, and bias mitigation. This paper proposes a risk-based approach for adopting AI/ML in drug development with a focus on human-led governance, ensuring the quality and reliability of data used for model development and clear documentation of criteria for developing and assessing models.

Current Adoption Trend and Future Outlook

The use of AI/ML has the potential to accelerate the drug development process and make clinical trials safer and more efficient. Most of the Pharma companies and Contract Research Organizations (CROs) are already adopting and building AIML based solutions for one or more use cases mentioned above and several such systems are already deployed in production. In 2021, the FDA received over 100 drug and biological product submissions with reference to AI/ML. These submissions spanned across different AI/ML use cases and drug therapeutic areas.

In the past, developing problem-specific models was an effort and time-consuming process. The advent of GenAI and the availability of ready-to-use pre-trained large language models (LLMs) now make it much easier to develop and deploy AI/ML-based solutions. Thus, AI/ML adoption in clinical trials is likely to become mainstream in the near future.

References:

1. Using Artificial Intelligence and Machine Learning in the Development of Drug and Biological Products
2. FDA Releases Two Discussion Papers to Spur Conversation about Artificial Intelligence and Machine Learning in Drug Development and Manufacturing
3. Future of AI & ML in Clinical Trials
4. AI Poised To Revolutionize Drug Development