AI holds the promise of revolutionising clinical trials, specifically by automating and optimising critical processes. The technology has the capability to significantly boost productivity, cut costs, and accelerate the overall timeline of clinical trials. 

In response to the need for cost-effective, accelerated, and inclusive clinical trials, there is a desire within the industry to embrace AI, however, adoption in the space is hindered by a lack of regulatory frameworks, in addition to potential data biases that limit the current scalability and more widespread adoption. 

‍

‍Background

As artificial intelligence technologies increasingly integrate into the healthcare and pharmaceutical sectors, the use of AI for clinical trials is garnering significant attention. The market is still relatively new, but is expected to witness a steady growth in the coming years. 

Generative AI for clinical trials is projected to be the fastest growing segment in pharma from 2022 to 2031. In 2023, the AI in clinical trials market was valued at USD 1.59 Billion in 2023 and projected to reach USD 6.55 Billion by 2030, growing at a CAGR of 22.4 % from 2023-2030. Within this space, the predominant market segment is clinical trial design, with the fastest-growing segment being outcome prediction. Moreover, interest is also growing in areas such as adverse event detection, sample size estimation, and patient recruitment.

‍

‍Applications of AI in clinical trials

‍

‍Improved Patient Recruitment and Retention

Patient recruitment and retention in clinical trials remains a significant challenge. Only 9% of Americans reported having ever been invited to participate in a clinical trial, and, of those who were invited, less than half reported participating, furthermore enrollment rates are less than 50% from pre-pandemic levels. AI accelerates and enhances patient recruitment by employing algorithms to gather subject information, evaluate potential participants, and analyse diverse data sources. It assists in effectively managing patients dispersed across various geographic locations and regions, thereby enhancing efficiency, quality, and retention. The company Quantiphi developed an AI-powered  multi-channel virtual agent for Wake Forest Baptist Medical Center’s clinical trials website, enabling efficient assistance to cancer patients in locating relevant clinical trials related to their condition. Quantiphi utilised, Dialogflow, Google’s natural language processing tool, to analyse the patient’s input as natural language, categorise the questions and bring together information on relevant clinical trials . The custom interface was then trained to provide clear answers to questions in real time. 

‍Reducing Study Timelines

The process of bringing a new drug to the market frequently spans 10 to 12 years, with the clinical trial stage alone averaging five to seven years. This extended timeline is primarily attributable to the challenges posed by manual effort, rework of data, and process inefficiencies. AI streamlines the complex process of constructing studies by automatically parsing protocols, generating electronic case report forms (eCRFs), and establishing databases within minutes. This results in a substantial reduction in study build timelines, improved accuracy, and streamlined data analysis, including the efficient generation of required tables, listings, and figures.

‍Outcome Prediction 

By leveraging AI to create predictive models rooted in patient characteristics and biomarkers, researchers can evaluate the response of individual patients to diverse interventions. This optimization of treatment efficiency and risk reduction is facilitated through AI's ability to develop insightful predictions. Insilico Medicine, a clinical-stage end-to-end generative AI drug discovery company, has developed InClinico for clinical trial outcome prediction. The transformer-based AI software platform brings together prediction engines that use generative AI and multimodal data, including omics, text, clinical trial design and small molecule properties. The company has recently demonstrated that the platform can predict the outcomes of Phase II to Phase III clinical trial success with a high degree of accuracy. 

‍External Control Arms

AI can enhance clinical trials by creating an external control arm, fostering patient-centric approaches, reducing enrollment timelines, and bolstering statistical power and result confidence. The company Unlearn, in collaboration with pharmaceutical sponsors, biotech firms, and academic institutions, is optimising a clinical trial software called TwinRCTs. The platform integrates AI, digital twins (virtual models mirroring physical objects), and novel statistical methods to enhance trial success rates with a smaller patient pool. In contrast to traditional trials, the AI model generates a digital twin for each patient based on historical control data, enabling predictions of disease progression in external cohorts. Accurate forecasts of treatment effects on primary and secondary endpoints are achieved by comparing the patient to their digital twin.

‍Results Sharing 

Generative AI can accelerate medical writing, efficiently summarising clinical trial results with high quality. It can analyse and synthesise research materials, presenting them in a clear format for various stakeholders, including team members, regulatory agencies, and review boards. Furthermore, it can simplify the preparation of regulatory submissions by automating the generation of documents and ensuring compliance and can generate sections of regulatory documents, such as clinical study reports, by extracting relevant information from databases and trial documents. Additionally, it facilitates the creation of plain language summaries for effective communication with patients, families, and the public, and supports translation into different languages. AI is already employed in various established applications for disseminating the findings and outcomes of clinical trials. An example of this is AI software company Yseop, whose content automation solution, Copilot, can be used to streamline the regulatory writing and submission process, utilising pre-trained LLM models. The tool has been previously used by Eli Lilly for drafting patient narratives.  

AI has the potential to impact many different aspects of the clinical trials process from trial design through to results sharing and dissemination, however, many of these applications are still theoretical. In the future, among other techniques, AI may be combined with smart devices, such as wearable sensor devices, to develop efficient, mobile, real-time, constant, and personalised patient surveillance systems that can monitor patients effectively during the trial period and reduce site visits. 

‍

‍Challenges

While AI solutions for clinical trials offer numerous advantages, it is crucial to address certain limitations in order to fully harness their maximum potential.

The effectiveness of AI in healthcare relies on the availability of high-quality data for pattern recognition and informed decision-making. In cases where the data is inconsistent, incomplete, or biased, AI may struggle to generate accurate predictions, leading to inefficiencies, potential risks associated with medications, and a failure to meet regulatory approval standards.

Another notable challenge involves safeguarding the privacy and security of patients' data. Given the daily handling of a substantial amount of sensitive information within the healthcare and pharmaceutical industry, there is a risk of unauthorised access and potential data breaches.

The incorporation of AI in clinical trials prompts concerns regarding patient safety and accountability in decision-making. The technology's ability to blur traditional roles underscores the need for clear and well-defined responsibility allocation.

‍

‍Conclusion

The potential for AI across the clinical trial process is endless. Although several tools have been developed, relatively few have been implemented in clinical trials, and even less have demonstrated tangible value such as a reduction in timelines or monetary gains. Furthermore, AI tools appear to be targeting individual aspects, with solutions yet to be developed that work across the clinical trials process as a whole. 

Despite this, as the technology continues to advance it’s likely to assume increasing responsibilities within clinical trials, as additional opportunities continue to arise, it is poised to bring further advantages to the clinical trials landscape.