AI in Clinical Trials: Improve Efficiency and Save Money

Clinical trials are the most important phase in the drug development process. They are used to assess the efficacy and safety of a drug, clinical approach, or therapy.

Currently, clinical trials are expensive, labor-intensive, and complex processes that are prone to errors and biases. Pharmaceutical companies have to individually select and recruit volunteers for clinical trials, which is time-consuming and expensive. 

However, artificial intelligence (AI) is transforming pharmaceutical companies conduct clinical trials, just as it is revolutionizing AI in drug discovery.

AI can be used to select clinical eligibility criteria, identify ideal clinical trial candidates, and reduce sample size requirements. This article discusses how AI-powered clinical trials can reduce costs and enhance efficiency.

How is AI transforming clinical trial processes?

AI can transform various aspects of clinical trials. Its aims include enhancing efficiency and reducing the costs of clinical trials. Here are some of the steps where AI in clinical trials can be used:

Patient selection and recruitment

AI algorithms can be used to sift through large databases of patient records. The desired criteria can be input into the algorithm, which can then provide a list of patients who match the eligibility criteria. This accelerates recruitment and ensures that the patient is suitable for the study.

Predictive patient recruitment

Pharma companies can use AI algorithms to sift through patient records and identify prospective clinical trial volunteers based on pre-defined data. Such integration will speed-up the patient recruitment process, reducing the timeline of clinical trials.

Furthermore, AI can also automate the initial phase of patient recruitment such as drafting initial permission letters and creating clinical trials decks for HCPs who interact with prospective clinical trials volunteers.

Patient retention

Patients often drop out of clinical trials, which poses challenges for researchers because it can create problems in data analysis. AI in clinical trials can be used to determine which candidates are likely to drop out.

By providing patients with regular surveys and feeding this information into the algorithm, AI can predict which candidate might be at risk of dropping out. AI can also provide solutions and strategies to improve patient retention.

Data management and analysis

During clinical trials, data is collected from various sources such as electronic health records, wearable gadgets, and medical devices. Collection, integration, and analysis of such vast databases are time-consuming. AI in clinical trials can handle complex databases and perform sophisticated analysis, which saves time and money.

Optimize trial design

AI simulations can be used to predict various scenarios and outcomes of clinical trials. Based on these results, the design of clinical trials can be optimized. Furthermore, potential challenges that may arise during trials can be predicted and solutions can be designed beforehand.

Trial monitoring automation

Clinical trials need to be thoroughly monitored throughout the period to ensure there are no anomalies. It involves constant and detailed data collection, analysis, and monitoring, for which employees are required, making clinical trials expensive.

However, many of these tasks can be performed by AI. For example, AI can analyze data and determine anomalies, which means conductors need not go through all the data.

AI can also automate repetitive compliance tasks such as data analysis, data extraction, report generation, submission, and quality checks, reducing human workload. This aligns with broader efforts in AI pharma regulatory compliance, helping companies stay audit-ready while cutting down on manual oversight.

Predictive modeling

AI can also be used to estimate the time and cost of clinical trials. The clinical trial design can be input into the algorithms along with additional parameters like number of participants. Based on this, the algorithm can provide an estimate of time and cost. The design can then be adapted to optimize both parameters.

Although AI in clinical trials is still at the nascent stage, it has already been integrated into some steps of clinical trials. Here are some key AI-powered clinical trial technologies.

Key AI-powered technologies used in clinical trials

Latest AI-powdered clinical trial software and technologies:

Trial Pathfinder is an open-source AI tool that uses electronic health records (EHRs) to simulate clinical trials. EHR data is integrated to understand how adjusting the inclusion criteria of the clinical trial can change the overall survival risk ratio. 

TwinRCTs is a clinical trial software that combines AI, novel statistical methods, and digital twins (a virtual model that can accurately reflect a physical object) to improve the success rate of clinical trials.

The model can predict disease progression in clinical trial patients using external cohorts. Digital twins are revolutionizing virtual trial simulations and predictive modeling in pharma. Learn more in our deep dive on digital twins in pharma.

Criteria2Query is an AI tool where users input the inclusion and exclusion criteria and the tool uses a web-based interface to find matching candidates in patient databases.

TrialGPT can be used to prompt a large language model to find appropriate clinical trials for a patient. A description of the patient is input, and the model determines whether the patient is eligible for any clinical trials, providing reasoning for its results.

Deep6 AI: Deep6 uses AI in clinical trials to match ongoing or up-coming studies with patients. It analyzes real-time unstructured clinical data to connect interested patients with relevant studies.

Trials.ai: Trails.ai employs an AI algorithm that streamlines the design and management of clinical trials. Researchers can analyze the cost and time associated with the design and optimize the same to ensure both remain within the desired limits.

IBM Watson Health: IBM Watson Health uses AI to analyze large databases of clinical trials to accelerate drug discovery.

These are some of the real-world applications of AI in clinical trials, complementing how AI in pharma marketing is enhancing patient engagement and outreach strategies. However, despite this adoption, the use of AI in clinical trials is facing some challenges.

Challenges facing AI in clinical trials

Challenge: Data is subject to stringent regulatory requirements. 

To predict trial outcomes, select correct patients for trials, and monitor patient health, AI algorithms need access to real-time, high-quality, and vast quantities of data.

However, considering the strict regulatory guidelines on data privacy and informed consent, these data may not always be available, which hampers the efficiency of the algorithm.

Challenge: AI integration into clinical trial workflow is expensive.

The implementation of AI in clinical trials is expensive. To integrate the robust infrastructure and computational resources (i.e., the database), companies have to hire experts.

This is a time and labour-consuming process, which turns out expensive. Furthermore, existing practices and systems also need to be integrated into AI to ensure it is more efficient than the traditional process.

Challenge: AI models are not transparent or interpretable.

Most AI models operate as black boxes, which means they do not provide any explanation for their recommendations or decisions. The lack of transparency and interpretability is problematic because, in clinical settings, each decision needs to have a logical reason.

Furthermore, AI models that do not provide reasoning or are not transparent may not be adopted because researchers will not trust the model.

Challenge: AI can carry forward biases from training data.

AI algorithms are trained on existing data, which may have biases. AI systems may inadvertently carry forward these biases, which may not apply to all patient populations, severely hampering their reliability. 

AI Ethical Considerations

Here are a few key ethical considerations for developing AI models for clinical trials:

The model must comply with all data regulations including the HIPAA and GDPR. These regulations exist to ensure data privacy and security, which is a primary concern for AI models. Additionally, anonymization processes must be implemented within the model to ensure privacy once launched.

Informed patient consent must always be obtained. Participants should be provided easy-to-understand and detailed information on how the data will be used before obtaining consent. This is also in agreement with all data regulations.

Responsibility for AI-recommended decisions must be established. The implementation of AI recommendations may lead to negative outcomes for clinical trials or patients. In such cases, it is necessary to determine who the responsibility lies with. Pushing the blame from one team to another will only create confusions and not provide any solutions. 

Validation needs to be thoroughly performed. While large variability and error ranges may be acceptable in other AI use cases, they aren’t for pharma because such errors may permanently impact patients’ lives. Hence, the reliability and fairness of the model must be thoroughly demonstrated across diverse populations and in real-world settings.

Transparency should be in-built. All recommendations and decisions from the model should be interpretable by clinicians and stakeholders. This transparency is often a challenge for most AI models, but it is crucial because only by understanding the reasoning behind a decision can it be determined whether the decision is right or wrong.

Risk mitigation should be a priority. In the pharmaceutical industry, ‘risk’ has a unique and significant definition: It implies permanent consequences for patients such as paralysis or death. Hence, risk of AI implementation in clinical trials needs to be thoroughly identified and mitigated before full-scale implementation to prevent any adverse events.

These considerations need to be addressed by all stakeholders during AI model building and training stages as they can severely impact whether the AI model will be approved by regulatory authorities and whether the model will provide the right answers.

Conclusion

The use of AI in clinical trials has a lot of potential, especially for enhancing effectiveness and reducing costs. AI-powered technologies can optimize data management and analysis, improve patient selection and retention, and optimize clinical trial designs.

To this end, several AI technologies have already been integrated by pharmaceutical companies. However, despite this integration, AI faces some challenges, primarily due to the strict data privacy laws, expensive integration requirements, and poor transparency.

As we move into the future, AI in clinical trials will have to overcome these and many other challenges to be widely adopted. However, considering its advantages, AI in clinical trials has huge potential. 

FAQs

How does AI accelerate clinical trials?

AI can accelerate clinical trials by automating repetitive tasks, identifying prospective participants, simplifying data management, optimizing trial design, and ensuring patient retention.

Examples of AI in clinical research?

AI can be used for new drug development, candidate identification, clinical trials automation, clinical trials design, and drug formulation. Pharma Now has published various articles detailing the use of AI in various branches of pharma. Read them to learn more

What are the ethical concerns with using AI in clinical trials?

Ethical concerns with using AI in clinical trials revolve around informed consent, data security, and patient privacy. Furthermore, AI may unknowingly adopt biases based on the data on which it is trained.

Are there any specific regulations on AI in clinical trials?

Yes, there are specific regulations imposed on AI. It is also subject to all data security, data privacy, ethical regulations, and guidelines applicable to all pharmaceutical companies.

Can AI be used for predicting clinical trial outcomes?

Yes, AI can be used for predicting outcomes. This allows researchers to make informed decisions during trials. 

References

https://www.nature.com/articles/d41586-024-00753-x

https://www.mckinsey.com/industries/life-sciences/our-insights/how-artificial-intelligence-can-power-clinical-development

https://www.nature.com/articles/s43856-023-00425-3

https://www.starmind.ai/blog/how-using-ai-in-clinical-trials-accelerates-drug-development

https://www.fda.gov/science-research/science-and-research-special-topics/artificial-intelligence-and-machine-learning-aiml-drug-development

https://www.linkedin.com/pulse/trial-pathfinder-ai-framework-systematically-evaluate-rahul-shirale/

https://www.linkedin.com/pulse/top-three-aiml-use-cases-largest-potential-clinical-trials-herax-fswqf/

https://www.nature.com/articles/d41586-024-00753-x#:~:text=Xiaoyan%20Wang%20notes%20that%20there,privacy%20or%20create%20security%20risks.

https://bmcmedicine.biomedcentral.com/articles/10.1186/s12916-019-1426-2