A Review On: AI Used in Clinical Trials

Abstract

The term “artificial intelligence” refers to a broad range of technologies that are currently being discussed and introduced in various clinical trials and areas of medicine, including patient requirements and enrollment, trial site selection, data collection, outcome assessment, safety monitoring, pharmacovigilance, and medical imaging. But there isn’t much information or literature on how AI methods can be used into clinical trial design. These articles will discuss various aspects of artificial intelligence (AI) being utilized in trials. Today, many AI methods are applied in clinical studies of new medications. Clinical trial safety and privacy ethical considerations of AI could reduce failure and speed up the drug development process.This study examines AI approaches that affect clinical trial results, concentrating on multimodal learning and prediction tools for trial embedding trails.

Keywords

Introduction

First of all, Artificial intelligence (AI) was long thought to as a science fiction concept that was only seen in movies. The use of AI and related technologies is growing in both business and society, and it is starting to show up in the healthcare industry. In addition to reforming administrative processes inside pharmaceutical companies, payers, and providers, these strategies have the potential to improve many areas of patient care. Numerous studies have already demonstrated that AI is capable of outperforming humans in crucial healthcare tasks.Such as the diagnosis of a sickness. Clinical studies are intended to evaluate a medicinal product’s safety and effectiveness in individual   Artificial intelligence (AI) was long thought to as a field of science fiction that was only seen in movies. The healthcare industry is starting to use artificial intelligence and related technologies, which are becoming more prevalent in both business and society.  Numerous facets of patient care as well as administrative processes in healthcare organizations, insurance companies, and pharmaceutical companies could be improved by these methods.  In crucial healthcare tasks like disease diagnosis, artificial intelligence can already outperform humans, according to a number of studies. The purpose of clinical trials is to evaluate the safety and effectiveness of a therapeutic product in treating a specific illness or condition in patients.On average, they take six to seven years, and they cost a lot of money.  The accepted approach for guaranteeing the efficacy and safety of new medications is still “linear and sequential” clinical trials.  Drugs that pass laboratory testing are often tested on humans.  Phase I, II, III, and IV trials are clinical studies that are conducted on humans after that.  A clinical study’s several phases each have a set of goals.Phase I studies generally demonstrate safety, phase II trials demonstrate efficacy, phase III trials demonstrate efficacy in contrast to current standard care, and phase IV trials show general benefits and dangers following medication approval.  As the medication development progresses through several phases, the number of participants increases.  During development, drugs that are shown to be hazardous or ineffective will not progress through all stages of the clinical trial phase. [1]

Different method of Ai used in clinical trials

Artificial Intelligence: Alternative Approaches The imitation of human intelligence processes by machines, including self-correction, learning, and reasoning, is known as artificial intelligence (AI).Mining for association rules Finding interesting correlations between variables in large databases using machine learning techniques is known as association rule mining. This facilitates the computer's ability to mimic the human brain's capacity to identify and abstract connections from novel, unclassified material.BMI, or brain-machine interface: The term “brain-machine interface” (BMI) describes a direct channel of communication between a wired or modified brain and an external device.  This technique is also known as the brain-computer interface (BCI), mind-machine interface (MMI), or direct neural interface (DNI).

Deep learning( DRL):Drawing inspiration from biological systems’ distributed communication nodes and information processing, deep learning (DL) is a class of machine learning techniques based on artificial neural networks. DL employs numerous layers to gradually extract higher level features from raw input. The term ‘deep’ in ‘deep learning’ describes the quantity of layers that are used to translate the data.

Deep reinforcement learning (DRL): In machine learning, deep reinforcement learning (DRL) is the field of reinforcement learning (RL), which focuses on creating software agents that can act in a given environment in order to maximize a concept known as cumulative reward. In order to develop effective algorithms for this task, DRL integrates the concepts of DL and RL.

Humane machine learning: When a person and a machine can communicate directly, it's called a human-machine interface (HMI). An example of a human–machine interface would be a synthetic system that can automatically understand and respond to spoken or written human language.

ML, or machine learning Machine learning (ML:) is the scientific study of algorithms that use sample data to build a mathematical model that can make decisions or predictions without specific instruction. A lot of people believe that machine learning is a branch of artificial intelligence.

Natural language processing (NLP): A branch of artificial intelligence called natural language processing (NLP) studies how computers interact with human (natural) languages, specifically how to teach computers to handle and evaluate vast volumes of natural language data. NLP incorporates elements of computational linguistics and computer science.

Optical character recognition (OCR): The goal of optical character recognition (OCR), a branch of artificial intelligence, pattern recognition, and computational vision research, is to electronically convert images of handwritten, typed, or printed text into machine-encoded text. This can be done from a scene photo, scanned document, or subtitle text superimposed on an image. [2]

Patient requirement and enrollment

The general phrase for participants undergoing recruitment, screening, and randomization is “patient enrollment in clinical trials.”  An effective way to conceptualize enrollment is as a funnel.  Following their identification as possible candidates.

-Patients undergo pre-screening,

-An initial site visit,

-Informed consent,

-Screening,

And, if successful, official enrollment as trial participants

When a patient completes screening and satisfies all eligibility requirements, they are officially enrolled.  Once they pass all screening tests, they are automatically enrolled because they have already completed an informed permission form to be screened. [3]

Fig one of barrier in patient requirement and enrollment.

Patient requirement and enrollment barrier

Complex Eligibility Criteria: Finding appropriate participants is becoming more difficult due to the growing complexity of inclusion and exclusion criteria in clinical studies. When trained on a variety of datasets, AI algorithms can assist in matching patients who meet these intricate requirements, guaranteeing their fitness for recruiting. [4]. Accessibility: Concerns around data breaches and privacy violations arise when sensitive patient data is accessed and used for AI-driven hiring. [5].

Integration with Current Systems: It can be difficult to integrate AI technologies with current trial management systems and electronic health record (EHR) systems. [6]

Site selection and monitoring by AI Data-Driven Decision Making: By analyzing massive information, AI systems can spot trends and forecast site performance, resulting in better site selection. Better Patient Recruiting: AI tailors patient recruiting tactics according to data patterns, resulting in more effective and focused recruitment initiatives.

Predictive Modeling: AI can forecast site performance by taking into account a number of variables, which aids in determining which sites are best for a given experiment. Predicting enrollment rates, data quality, and procedure adherence are all part of this. [7]

Decreased Costs and Time: AI expedites the site selection procedure, saving time and money on manual data analysis and site assessment. [8]

Personalized Recruitment: By using AI to evaluate patient data, recruitment techniques can be improved and diversity can be increased by finding possible volunteers that fit the intricate inclusion and exclusion criteria for a trial. [9]

Automatic Data Gathering: The need for frequent clinic visits is decreased by using AI algorithms to analyze data from wearable devices (such as smartwatches and biosensors) to track vital signs, activity levels, and other patient-reported outcomes in real-time. [10]

Decreased Manual Effort: AI relieves clinical personnel of a lot of data administration chores, freeing up their time for other important duties. [11]

Transparency and Explainability: Ensuring the transparency and explainability of AI algorithms utilized in clinical trials is essential, especially with regard to their decision-making procedures. [12]

Managing and Analyzing Data: Clinical trial data administration is being revolutionized by AI, which speeds up timelines, increases accuracy, and automates operations. Large datasets, data extraction from several sources, mistake detection and correction, and insightful analysis all possible with AI-powered systems. Increased effectiveness, lower expenses, and eventually quicker and more trustworthy clinical studies are the results of this. [13]

Pharmacovigilance and safety monitoring

 In order to monitor adverse drug reactions (ADRs) and guarantee drug safety, pharmacovigilance is essential.  Conventional approaches are unreliable and slow, while artificial intelligence (AI) increases accuracy and efficiency in handling growing data complexity through automation and sophisticated analytics. [14]  With advantages including improved adverse event detection, data-driven risk prediction, and optimized drug development, artificial intelligence (AI) has the potential to revolutionize pharmacovigilance.  Generative artificial intelligence has applications in a range of disease states by simplifying the processing and analysis of large datasets.  Pharmacovigilance processes can be streamlined and safety-related data can be evaluated more effectively with the use of machine learning and automation in this area. [15]

The use of AI in pharmacovigilance Automatic adverse events reporting and case processing

1. Case handling and adverse event (AE) reporting that is automated The collecting and processing of safety data from several sources, including unstructured data, can be automated by AI-driven solutions, which can also transform the data into coded medical information and defined formats. Additionally, these systems are able to detect duplicate reports. Improved determination of causation The workload for human experts is lessened and real-world evidence can be integrated. Enhanced causality

 2. when AI-based methods, such as machine learning and natural language processing, evaluate various datasets to ascertain the possibility of a causal relationship between a medication and an adverse event. Bayesian networks defined by experts have been used to guarantee reliable evaluations. [16]

3.Signal detection and analysis AI tools use ML and predictive analytics for proactive safety signal identification, continuously monitoring data from sources like EHRs and social media. Data mining techniques are employed to classify drug-AE relationships, and AI tools can also assist in literature searches to evaluate relevant articles.[17]

Outcome Assessment

AI is transforming clinical trial outcome assessments by enabling faster, more efficient, and patient-centric data analysis. AI, particularly machine learning (ML), can analyze large, complex datasets from various sources (e.g., electronic health records, wearables, imaging) to predict disease progression, treatment response, and patient outcomes. This allows for more personalized care, improved patient selection, and faster interventions. [18]Enhanced Data Analysis: To find trends and forecast results, AI algorithms may examine enormous volumes of data from a variety of sources, such as clinical trial databases, medical literature, and electronic health records. [19] Natural language processing Clinical notes and patient reports are examples of unstructured text from which Natural Language Processing (NLP) can extract valuable information, allowing for a more thorough comprehension of patient experiences. [20] Personalized Assessment AI-powered technologies have the ability to customize outcome evaluations according to the unique traits and preferences of each patient, producing data that is more pertinent and useful. [21] Personalized Healthcare By determining which patients will benefit most from a certain treatment, AI helps to design tailored health strategies. [22]AI in clinical trial protocol design: Important uses for AI in clinical trial protocol design Protocol Optimization: AI can find the best trial designs, including patient selection standards, treatment strategies, and data gathering techniques, by analyzing real-world evidence, patient records, and trial data from the past. [23]. Enhanced patient centric design Improved design that is focused on patients Trial protocols can be modified to be more patient-centric by utilizing AI, which could improve study protocol adherence and participant engagement. AI can assist in creating protocol-related documents that are easier to access and utilize, proactively addressing any issues raised by possible participants. [24].

Ethical considerations

A consideration of ethics Social and clinical value, scientific validity, participant selection, risk-benefit ratios, and informed consent are all ethical factors to be taken into account in AI-driven clinical trials. Trials using AI must show a definite benefit, particularly in lowering health inequalities, and be scientifically valid using techniques like randomized controlled trials. Transparent, customized informed consent, balanced risk-benefit profiles, and equitable participant selection are also essential. Fairness and bias Algorithmic Bias: If AI algorithms are trained on non-diverse datasets, they may reinforce or worsen preexisting prejudices, producing unfair or incorrect treatment recommendations, diagnoses, and predictions for particular patient groups. Inclusivity and Equity: It’s critical to guarantee that everyone, irrespective of socioeconomic background, race, or other characteristics, has fair access to AI-driven healthcare solutions, representative datasets, and fair participant selection. Effect on Inequalities in Health: Because AI has the potential to exacerbate already-existing health inequities, especially among marginalized people, proactive measures are needed to guarantee fair access and advantages. [25]

CONCLUSION

Artificial intelligence-based clinical trials are a new and exciting field of research. Future drug development could be completely transformed by AI, which could also open the door to a new paradigm of sustainable, long-term medical research. The comprehensive strategy to integrating AI into drug research and approval covers every stage of a medicine's lifetime. From target identification to medication clinical trials, AI can streamline the process. The majority of the time and money spent on drug research goes into clinical trials for new medicines, and artificial intelligence (AI) has been used to enhance trial design, patient and dose selection, patient adherence, trial monitoring, and endpoint analysis.AI technology must be morally sound, reproducible, scalable, and intelligible for both end users and regulatory agencies. In this sense, AI-enabled methods will open up a range of possibilities for clinical research, potentially transforming the direction of subsequent investigations. The benefits of AI tools in the healthcare sector won't be completely realized for another five to eight years. The sector needs particular legislation, clear assessment standards, and a positive mindset because widespread AI adoption is currently hampered by certain issues.

ACKNOWLEDGEMENT

The authors would like to express their sincere gratitude to all researchers, clinicians, and data scientists whose pioneering work in artificial intelligence and clinical trial methodology made this review possible. We also acknowledge the contributions of open-access databases, journals, and online platforms that provided invaluable information and insights. Special thanks are extended to colleagues and mentors who offered guidance, feedback, and encouragement throughout the preparation of this manuscript.

REFERENCES

1. Juneja A, Chaturvedi S, Shrivastav S, Dhar H, Beri A . A Artificial Intelligence Revolutionizing Clinical Trials Process (Dec,2022, 11(12):5625-5642.
2. Harrer S, Shah P, Antony B, Hu J. Artificial intelligence for clinical trial design. Trends in pharmacological sciences. 2019 Aug 1;40(8):577-91.
3. U.S. Department of Health and Human Services. Recruitment and enrollment in clinical trials [Internet]. Bethesda (MD): National Institutes of Health; [cited 2025 Oct 4]. Available from: https://www.nih.gov
4. Calado G, Askin S, Burkhalter D, El Dakrouni S. Clinical studies using artificial intelligence: potential and difficulties. Health Technol (Berl). 2023;13(2):203-13.
5. MedTech Insights. AI for clinical trial recruitment: transforming methods for enrolling patients. MedTech Perspect. 2023 Jun 15.
6. Liang L, Hu G, Zhong Z, Jiang Z, Yang C, Lu X. A scoping study on the application of AI to improve clinical trial enrollment and retention. J Am Med Inform Assoc. 2024;31(11):2749-59.
7. Mak KK, Pichika MR. Current state and potential future use of artificial intelligence in medication development. Drug Discov Today. 2019;24(3):773-80.
8. Willard JO, Crowther J, Dobrindt A. Three AI capabilities are highlighted for improving clinical trial site performance [Internet]. Consulting by IBM. 2023 Aug 7 [cited 2025 Oct 4]. Available from: https://www.ibm.com
9. Askin S, Burkhalter D, Calado G, El Dakrouni S. Clinical trials using artificial intelligence: prospects and difficulties. Health Technol (Berl). 2023;13(2):203-13.
10. Wood T. How can clinical trials employ AI? [Internet]. Quick Data Science. 2024 Aug 9 [cited 2025 Oct 4]. Available from: https://fastdatascience.com/ai-in-pharma/ai-in-clinical-trials/
11. Jeeva Clinical Trials. AI or Bye: using cutting-edge technology to revolutionize clinical trials [Internet]. 2024 Mar 12 [cited 2025 Oct 4]. Available from: https://www.jeevatrials.com
12. Foote HP, Hong C, Anwar M, Borentain M, Bugin K, Dreyer N, Fessel J, Goyal N, Hanger M, Hernandez AF, Hornik CP, Jackman JG, Lindsay AC, Matheny MEE, Stockbridge N, Embi P, Ozer K, Seidel JJ, Lindsell CJ. Using generative AI in clinical research and other fields: possibilities, difficulties, and solutions. JACC Adv. 2025;4(3):101593.
13. Benefits and use cases of machine learning and artificial intelligence in clinical trial design [Internet]. 2025 Jun 30 [cited 2025 Oct 4]. Available from: [insert URL if available]
14. Algarvio RC, Ferreira-da-Silva R, Ribeiro I, Rodrigues PP, Conceição J. A literature study and practical use of a Bayesian network tool created by specialists on artificial intelligence in pharmacovigilance. Int J Clin Pharm. 2025;47:932-44.
15. Gupta R, Misra HP. Using generative AI to improve pharmacovigilance and drug safety. Clin Exp Pharmacol. 2025;20(2):89-97.
16. Barbieri MA, Heckmann NS, Papoutsi DG, Carnovale C, Cocco M, Battini V. An outline of the artificial intelligence-related technologies, applications, research, and regulatory landscape in pharmacovigilance signal processing. [Journal/Source info if available]. 2025.
17. .Algarvio RC, Ferreira-da-Silva R, Rodrigues PP, Ribeiro I, Conceição J. Pharmacovigilance with artificial intelligence: a narrative review and hands-on experience. [Journal/Source info if available]. 2025.
18. Marksman Healthcare. The function of artificial intelligence (AI) in clinical result evaluation: digital biomarkers, patient-centered technology, clinical outcomes evaluation, and AI in healthcare [Internet]. Healthcare by MarksMan. [cited 2025 Oct 4]. Available from: https://www.marksmanhealthcare.com
19. Albadawy M, Khalifa M. Examining important areas and crucial features of artificial intelligence for clinical prediction. Biomed Comput Methods Programs Update. 2024;5:100148.
20. Li S, Zhu J, Haris P, Qian L, Yang Y, Lu X. A comprehensive analysis of how artificial intelligence can be used to improve clinical trial outcome prediction. J Drug Discov. 2025;30(4):104332.
21. Chopra H, Priyanka, Dhama K, Annu, Shin DK, Munjal K, Emran TB. Clinical trial revolution: AI’s contribution to speeding up medical advancements. [Journal/Source info if available]. [Year].
22. Medio. Future developments are shaped by AI in clinical trials [Internet]. Medrio. 2025 Jan 13 [cited 2025 Oct 4]. Available from: https://www.medrio.com
23. Jordan ER, Jahriss L, Kasi PM. Putting in place an artificial intelligence-powered internet search engine to help people find and understand trustworthy tumor gastrointestinal clinical trials [Internet]. [cited 2025 Oct 4]. Available from: [insert URL]
24. Chenchala PK, Devaguptapu B, Choppadandi A, Kaur J, Naguri S, Saoji R, Tilala MH. A careful examination of the moral dilemmas raised by the use of AI and machine learning in healthcare. Cureus. 2024;16(6):e62443