AI in Clinical Trials: How Are AI and ML Impacting the Clinical Trial Process?

Using AI to accelerate clinical trials: Potential breakthroughs and associated challenges to address

Artificial intelligence (AI) is quickly finding its place in the clinical research industry; approximately 1573 trials using AI in some way were conducted between 2016 and 2022.1 Many stakeholders in the industry believe AI holds huge potential for improving almost all aspects of clinical research study design and conduct.

Furthermore, artificial intelligence in pharmaceutical industry R&D could help guide drug development efforts in directions that will result in more successful trials and the best outcomes for patients. Nonetheless, there are prominent concerns and challenges associated with the use of ML and AI in clinical trials that need to be addressed, ranging from data privacy and ethical concerns to potential bias.

In this article, we investigate specific uses as well as some of the benefits and challenges of these innovative algorithms to discover how AI and ML are changing the clinical research landscape. 

Before continuing, it is important to explain how machine learning (ML) algorithms differ from AI. Machine learning is an extension of AI that follows the same basic principles of being trained on data to identify patterns, which are then used to provide information or complete complex tasks.2 However, the similarities more or less end there. 

AI algorithms essentially aim to mimic human thinking; if a mistake is made, it is adjusted and improved by developers, who reprogram the algorithm to take into account any new variables. Therefore, there is human involvement in making the AI algorithm more accurate.3

On the other hand, ML algorithms have built-in functionality to self-check and improve when a mistake is made, thus requiring less developer involvement. So, in a way, the ML system is developing its own intelligence with its inherent learning ability, which is what sets ML technology apart from AI.2,3 

How are AI and ML used in clinical trials? 

In the realm of clinical research, many aspects of both design and operations can potentially be improved with AI and machine learning. Clinical trials can leverage AI and ML to facilitate, accelerate, or follow new directions in aspects such as:4,5,6

AI can process electronic health records (EHRs) in a fraction of the time it would take to manually review, sort, and compile data. EHRs can be mined to sort prospective patients and identify patterns which could then be used to define eligibility criteria or to identify qualifying populations to create cohorts for a specific clinical trial. 

AI can help streamline the clinical trial recruitment process by rapidly screening databases to identify eligible patients, for example by comparing electronic health records (EHRs) against inclusion and exclusion criteria. Furthermore, AI technology can also be used to notify qualifying patients about clinical trials that they would be eligible to apply for. 

AI in clinical trials makes it more feasible to analyze and use new data sources and real-world data, such as healthcare facility patient databases, EHRs, and insurance claim records.  

Automated data processing and deeper insights

AI or ML-enabled tools can be set up to automate aspects of data collection and data processing, for example by intelligently populating reports and performing automatic validation checks. This also allows for faster and deeper insights into the data, overall helping trials move faster through their timelines.  

Generation of content that adheres to regulatory frameworks 

Artificial intelligence tools have already shown strong capability to generate various types of content. If trained appropriately on regulatory frameworks, review board submissions, and legal documentation, AI and ML could even write compliant first drafts of protocols which research teams can then edit and polish to reach a final version. The same logic can be applied to the creation of recruitment and promotional materials, and even reports for sponsors. Overall, this application of AI holds the potential to greatly minimize the overall time spent on paperwork and writing.

Patient information and perspectives from sources such as online support groups, patient advocacy organizations, and follow-up feedback from past trials could be analyzed to assist sponsors and CROs in designing more patient-centric clinical trials and protocols.7

With the enormous amounts of data currently available, AI tools could be used to assist analyses of trial data and healthcare trends (similar to health economics and outcomes research; HEOR) to highlight past successes, gaps in treatment and care, and issues or areas of concern, which can support more strategic decision-making to improve drug development efforts, clinical trial operations, treatment indications, and healthcare guidelines and policy. 

AI and machine learning tools are excelling rapidly in image processing, and can be trained to recognize abnormalities in medical images, such as CT scans and MRIs, often with much greater sensitivity than that allowed by the human eye. This can assist clinical researchers in rapidly and accurately assessing health outcomes and endpoints that involve medical imaging.8

AI and ML tools have also shown promise in predictive modeling, combining various health indicators and other factors to make clinical predictions about the outcomes of different interventions (such as mortality, adverse events, recurrence rates, etc.). If applied strategically in certain study designs, this could potentially reduce the need for control arms or placebo groups. 

Identifying the most important outcomes for outcomes-based research 

AI and ML can also be trained to identify outcomes of highest priority, i.e., the health measures that are most relevant to patients and healthcare providers. This ties in with outcomes-based research, an approach that guides sponsors and researchers to design studies that work towards practical results and tangible benefits for patients. 

What are some challenges and disadvantages of AI in clinical trials?

As we have just seen, there are many promising ways AI and machine learning can improve clinical research. Nonetheless, the use of these technologies carries several concerns that are important to address. 

Since AI and ML require enormous volumes of data in order to learn comprehensively and be effective, data management becomes a primary consideration.4 Currently, there is no generalized standard for EHRs in terms of data formats and file types. In general, there is very wide variability in data sources, which makes it significantly more difficult to develop an algorithm that can process all of them coherently.9 This limits the current utility of the vast amounts of data, at least until agreements are reached on standardization. An ultimate goal, which would facilitate batch data analysis over huge datasets, is to develop a universal standard for EHRs. However, this is a complicated undertaking considering the heterogeneity in regulations and practices at the organizational, local, regional, state, and global levels.

Moreover, there is still a significant reliance on handwriting in patient records. AI and natural language processing still have a ways to go before they are sophisticated enough to read the wide variability in  handwriting and digitize and analyze handwritten medical files. 

Another challenge relates to potential bias arising from the datasets which are currently available. The capabilities of AI and ML systems depend on the data on which they are trained, so if they are fed with selective or specific data, it can lead to significant limitations or bias in the resultant applications. For example, if health records are not available for a specific region, the local ethnic groups and their unique genetic and demographic features may not be captured at all in that algorithm. In that case, extreme care should be taken in generalizing any findings to such populations, or they should be excluded from applications of the system altogether. Under the same logic, depending on the variability in the data utilized, there is room for significant bias to be present. Care should be taken to explicitly acknowledge and account for limitations of these technologies as they do not represent universal sources of truth. Another potential source of bias lies in assessing and classifying/quantifying subjective data, such as patient-reported symptoms or open-ended questions; such data sources require clear guidelines for their interpretation according to predefined rules.

What are the ethical considerations of AI in clinical trials?

AI in clinical trials also raises ethical concerns, particularly regarding data privacy and protection of participants. Regardless of the specific functions the AI or ML tool is tasked with completing, it must operate within the bounds of all applicable regulatory and ethical standards and frameworks, further stipulating that:10,11

Developing and implementing intelligent technologies that can comply with all of these considerations requires a great deal of quality control and testing. Further, they should be scalable to support adaptability to trials of different design and extent.

Given the relative objectivity (or indifference) of intelligent systems and their lack of a ‘moral compass,’ some patients may not feel comfortable with such technologies being employed in their healthcare. It is critical to retain a human element in controlling and overseeing uses of AI and ML, in order to ensure that operations comply with ethical principles, to uphold patient safety and data security at all times, and to control for known potential bias, in addition to maintaining the perceived comfort and familiarity of human involvement in patients’ health journeys.12

What is the future of AI in clinical trials?

AI seems to have found various applications in clinical research and healthcare, but how the future of AI in clinical trials will look will depend largely on further developments, both in technologies and in ethical and regulatory landscapes.

There are new horizons emerging in clinical research that are closely related to advanced analytics capabilities, powered by or in tandem with AI and ML. These include concepts such as personalized medicine and precision medicine, remote and automated data collection (i.e., wearable devices and remote health monitoring), and analysis of huge swaths of real-world data (RWD), for example in the mining of EHRs using natural language processing. These advances are opening doors to entirely new clinical trial models, and enabling the use of much larger pools of data as well as much more precision and specificity in the conclusions that can be drawn. However, we are at a point where it is crucial to guide these advances in the right direction, making sure that research efforts and policy changes lead to tangible and practical benefits for patient health and well-being.

Below, we investigate this topic further by exploring two questions: whether or not AI can replace researchers, and whether or not AI can make better clinical decisions than humans.   

Many prominent clinical research companies are integrating AI and machine learning into their research methodologies and digital solutions, even offering specific clinical AI services. A few examples include:

IQVIA is a US-based multinational that provides intelligent solutions to pharmaceutical, biotech, and medical companies conducting clinical trials.15 They explicitly leverage AI in some of their services, and have published a whitepaper on the use of AI in clinical development.14

ICON PLC a multinational CRO based in Ireland that provides consulting and intelligent clinical development services to the pharmaceutical and clinical research industry.16

Vial is a CRO that aids sponsors and researchers with clinical research services as well as cutting-edge tools leveraging the latest technological advances, including AI.17

[1] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9602501/ 

[2] https://azure.microsoft.com/en-us/resources/cloud-computing-dictionary/artificial-intelligence-vs-machine-learning/ 

[3] https://pubmed.ncbi.nlm.nih.gov/35135685/ 

[4] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8011519/ 

[5] https://link.springer.com/article/10.1007/s12553-023-00738-2 

[6] https://hai.stanford.edu/news/ai-expands-reach-clinical-trials-broadening-access-more-women-minority-and-older-patients 

[7] https://pubmed.ncbi.nlm.nih.gov/36292444/ 

[8] https://bmcmedimaging.biomedcentral.com/articles/10.1186/s12880-022-00753-1 

[9] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9908503/ 

[10] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8963864/ 

[11] https://www.sciencedirect.com/science/article/pii/B9780128184387000125 

[12] https://www.frontiersin.org/articles/10.3389/fsurg.2022.862322/full 

[13] https://www.tandfonline.com/doi/full/10.1080/0773937.2020.1801562 

[14] https://www.iqvia.com/-/media/iqvia/pdfs/library/white-papers/ai-in-clinical-development.pdf 

[17] https://vial.com/blog/articles/cros-and-artificial-intelligence-how-ai-is-changing-clinical-research/ 

AI in Clinical Trials: How Are AI and ML Impacting the Clinical Trial Process?

Other Trials to Consider

Active HD-tDCS

Magnetic Resonance Imaging using 7 Tesla MRI

PRA023

DystoniaNet

NeuViCare application

LyssnCBT

Adaptive Music Intervention

e-CBT + Phone Call + Pharmacotherapy

NT-proBNP

OCU410ST

Popular Categories

Levitra Clinical Trials

Aortic Stenosis Clinical Trials 2023

Autism Clinical Trials 2024

Diabetic Retinopathy Clinical Trials 2023

Colorectal Cancer Clinical Trials in Baltimore, MD

Cataract Clinical Trials 2024

Yeast Infection Clinical Trials 2024

Rhabdomyosarcoma Clinical Trials 2024

Paid Clinical Trials in Nevada

Dystonia Clinical Trials 2023

Popular Guides

Learn More About These Treatments

Procalamine

Blephamide

Benzamycin

Permapen

Minirin

Phrenilin Forte

Delstrigo

Estradiol Cypionate

Winlevi

Reserpine