As early as the 1940s, computer scientists have been working on making machines as intelligent as human beings, starting with Theseus, a remote-controlled mouse built in 1950 that was able to find its way out of a labyrinth and could remember its course. Since then, artificial intelligence (AI) systems have seen an unprecedented growth rate. Just 15 years ago, there was no machine that could reliably simulate language or recognize images as well as a human. However, in the last few years, these systems have steadily become more advanced and are now even surpassing human performance in some areas. No fields have seen as rapid an expansion in the number of opportunities provided by AI research as the medical and pharmaceutical industry have.

Unlike computer technology, which has become more easily accessible in terms of price and usability over time, clinical trials have only become more complex and expensive for sponsors and contract research organizations (CROs). In 2022, new drug approvals decreased by 25% in the United States (US) by the Food and Drug Administration (FDA), the lowest numbers seen since 2016. However, the advent of AI research in the drug development space may help researchers address the challenges modern clinical trials are faced with.

In this article, we will explore what role AI research has begun to play in the world of clinical trials, highlighting its benefits, applications, and potential future developments. Continue reading to learn more!

 

Enhancing Study Design in Clinical Trials

The first step of a clinical trial’s lifecycle is its study design; AI/machine learning (ML) algorithms can streamline this process in multiple ways. For example, ML algorithms can be combined with existing statistical approaches to determine the most appropriate sample size needed in a study. This way, researchers gain statistical robustness by maximizing the statistical power of a study, while the protocol also considers the ethics perspective by minimizing unnecessary patient exposure.

This Nature article highlights the emerging use of deep learning (DL) models to help researchers collate information from clinical trial publication abstracts or records from ClinicalTrials.gov and gain key study design insights. By extracting safety, efficacy, and other characteristics from these sources, sponsors can determine how previous researchers have designed trials or how newer digital tools (e.g., wearable devices) are being implemented, and what the outcomes were. These insights allow lessons to be learned from existing clinical trials, increasing the chance of improvements in new protocol designs.

 

The Potential Role of AI Research Models in Patient Recruitment

As for patient recruitment, it is no secret this is the biggest hurdle any sponsor or CRO faces when starting a new clinical trial. With AI applications, predictive analyses can be done to determine which patient populations are most likely to benefit from a particular treatment. With a more defined target study population in mind, recruitment strategies can be adapted with similar specificity. Another key potential application for AI is leveraging its ability to process massive volumes of data to rapidly review electronic health records (EHR). Identifying potential participants is an incredibly time-consuming process for site staff that often takes away valuable time and resources from other critical tasks in clinical trials, but AI is providing a vital solution by introducing automation capabilities and promoting efficient resource allocation.

For example, Deep6 AI uses AI and natural language processing (NLP) to find precise matches between eligible potential patients and clinical trials. The company offers a user-friendly ecosystem with ICD-10 medical codes, physician notes, lab results, and more to help sponsors with recruitment. The platform even allows a study’s inclusion and exclusion criteria to be included in the search parameters, with real-time updates on which criteria are impacting patient totals or partial matches. Read more about how AI is improving patient eligibility criteria and patient recruitment here!

 

Embracing AI Research in Drug Development at the Regulatory Level

In 2023, the US FDA released two discussion papers describing the use of AI/ML in drug and biologic development and manufacturing, and the development of medical devices. In 2021, the FDA received more than 100 drug and biologic applications which included AI/ML components, demonstrating the importance of recognizing the role of AI by regulatory bodies. These agencies (e.g., the FDA, the European Medicines Agency (EMA), and others worldwide) function to help ensure that approved drugs are safe and effective, while also promoting innovations in development. The first discussion paper released last year emphasizes the FDA’s support of increasing implementation of AI/ML in clinical research, but it also highlights the need for continued human involvement. The US-based regulatory agency recommends adopting a risk-based approach to evaluate and manage AI/ML, as well as the validity of datasets used to train ML or DL models. These approaches were presented as a way to prevent and address potential algorithmic discrimination with these technologies and maintain equity when using AI/ML techniques. More detailed insights provided by the FDA can be found in the following discussion paper, “Using Artificial Intelligence and Machine Learning in the Development of Drug and Biological Products.”

The FDA’s Center for Drug Evaluation and Research (CDER) also released a second discussion paper, “Artificial Intelligence in Drug Manufacturing”. The report recognized that AI technologies have a crucial role in manufacturing processes, such as enhancing controls, spotting early warning alerts, and minimizing product losses. CDER is also taking steps to hold workshops and provide consultation to relevant regulatory stakeholders to better understand how these technologies can be safely implemented into clinical research and drug development.

 

Improving Data Analytics and Safety Monitoring in Clinical Trials

Clinical trials generate enormous amounts of complex data, including clinical measurements, imaging results, and patient-reported outcomes, which is where AI capabilities come in to streamline the analysis of such large datasets. For example, DL models can be trained to detect subtle clinically relevant changes in medical images with greater accuracy than human assessors, potentially identifying early signs of disease progression or treatment response. In fact, this application is specifically promoting better screening programs in the field of ophthalmology to diagnose diseases like diabetic retinopathy!

The ability to monitor incoming drug efficacy and safety results in real-time, AI/ML techniques can be used to run more advanced adaptive clinical trial designs. As a refresher, adaptive clinical trials are a type of study that can use interim study results to inform protocol modifications while the trial is still ongoing, reducing inefficient use of valuable time and resources. Real-time monitoring of incoming patient data can also help sponsors, CROs, and principal investigators (PI) make informed medical decisions in response to anomalies or adverse events.

 

Medidata Clinical Data Studio: Recent Innovation in AI-Based Data Analytics

Did you know TFS HealthScience became the first CRO to adopt Medidata Clinical Data Studio (formerly called Medidata Detect) for its clients in 2023? Medidata is a well-recognized industry leader in data management innovations for clinical trials and its Clinical Data Studio is no exception. It provides a comprehensive data surveillance and risk monitoring solution that improves data quality and promotes patient safety in clinical trials for global CROs like TFS, and most recently, Worldwide Clinical Trials.

Using a unified data platform that pulls from various diverse sources of patient data, Medidata’s Clinical Data Studio leverages AI models to detect possible data issues and safety signals more effectively. Its workspace includes AI-assisted data reconciliation, self-serve data listings, anomaly detection and risk-based quality management, all of which result in accelerating data review times by as much as 80% for sponsors and CROs. Read more here about TFS CRO’s partnership with Medidata or learn about the Clinical Data Studio in this article!

 

Challenges and Ethical Considerations for AI Research

Despite these numerous benefits that AI integration into clinical research presents, it also comes with unique challenges and ethical considerations. Primarily, many researchers are calling attention to the potential for bias in AI algorithms, referred to as “algorithmic discrimination” by the FDA, which occurs when a model is trained using unrepresentative data or has a flawed design. As a result, these biases could favor one category of people over another in a study, assuming the training dataset was not sufficiently diverse in terms of patient groups, exacerbating existing health disparities and skewing trial results.

Another concern is the lack of inherent transparency and interpretability of AI models used in clinical research. Regulatory bodies and healthcare professionals should reasonably be able to understand how and why an AI/ML model has arrived at certain conclusions. The FDA also highlighted concerns about data privacy and security in their 2023 discussion paper, given the sensitive nature of patient health information used in clinical trials. Therefore, researchers using AI in their clinical trial designs and applications have the difficult task of balancing the need for data sharing for research purposes while protecting patient confidentiality at all times. Lastly, the rapid boom of AI use in the pharmaceutical industry calls for updates to the regulatory framework to ensure its use is governed carefully to maintain patient and drug safety in clinical research.

 

Conclusion

In conclusion, AI research has quickly established a critical role for itself in the realm of clinical research, introducing new opportunities to make the drug development process faster, more efficient, and cost-effective. From optimizing trial designs and enhancing patient recruitment to revolutionizing data analysis and safety monitoring, sponsors and CROs are increasingly implementing AI/ML technologies to streamline their clinical trials. Although there are some challenges recognized by regulatory agencies like the FDA, including potential concerns with bias, transparency, and ethical compliance, these techniques still hold significant promise for patients and researchers.

With new advancements in ML/DL models in the future, we may see new applications in clinical research, such as personalized and adaptive clinical trial designs or integration with blockchain or Internet-of-Things (IoT) devices. As the field grows, AI research can be expected to continue to play a central role in improving clinical trials and promoting better patient health outcomes.

 

About TFS HealthScience CRO

TFS HealthScience is a global CRO that supports biotechnology and pharmaceutical companies throughout their entire clinical development journey. In partnership with customers, we build solution-driven teams working for a healthier future. As a trusted CRO partner throughout the entire clinical development journey, we understand the importance of providing essential and diverse services to streamline clinical trials for our clients. Visit our website to learn more about the solutions TFS CRO can offer for your next study or connect with a TFS representative today!