Is AI the Future of Clinical Trials?

Artificial Intelligence Clinical Trials main

Clinical Trials are the mandatory path for developing and bringing a new drug or vaccine to the market.

Unfortunately, according to a study conducted by MIT, 86 percent of the drugs will fail during this process. This very high failure rate not only has consequences on the Pharmaceutical companies’ bottom line, but it precludes potentially safe and efficacious drugs from reaching patients that could benefit from them.

Some of the reasons that explain the failures of more clinical trials include:

  • Lack of efficacy
  • Lack of Safety
  • Poorly chosen endpoints
  • Non optimal study design and targetted population

Recruitment is one of the main bottlenecks, is time-consuming, and very expensive. According to Chunhua Weng from Columbia University (New York), “Recruitment is the number one barrier to clinical research.”

With its ability to analyze vast amounts of structured and unstructured data, Artificial Intelligence may be on the verge of changing how clinical trials are conducted and help in dramatically improving the success rate.

AI has the potential to be used in all stages of drug development, including the identification of new targets, design of the molecules, improvement of the clinical trial design, planning, execution, and finally for pharmacovigilance.

According to Dr. Bhatt, who recently published an article in the Perspective in Clinical Research journal, clinical trials can benefit from AI at different levels.

Reducing Patient Heterogeneity

Reducing patient heterogeneity is often used to demonstrate the efficacy of a new treatment compared to the control/placebo group. However, the benefit of such an approach has been recently challenged as AI is one of the key approaches to defining a relevant subset of patients.

Identifying compounds for the age of precision medicine

Auransa is based in Palo Alto, California. The pharmaceutical company relies on the SMarTR AI engine to predict the effects of a particular drug candidate on patients’ subpopulations.

Instead of testing thousands of drug candidates, SMarTR runs in silico computational models using clinical, genetic, and demographic information to characterize patients’ subtypes. As a result, the company is at the forefront of what precision medicine could be in the future.

The initial computational states identify compounds in a target agnostic way, conducting clinical trials only in patients with the highest chances of benefit. As stated by Pek Lum, the company COO and co-founder, the goal is not “to get drugs into the clinic faster, but to target responders.”

AI can thus improve patients’ selection and make them more prone to respond positively to the potential treatment. Yet, electronic phenotyping is at the core of the future of medicine.

The future of drug design and medicine will not be a “one drug suits all” approach anymore. Instead, especially with genotyping becoming more affordable, it is expected that drugs will be 3D-printed on-demand to better answer patients’ specific needs based on their genetic profile.

Using AI for designing more inclusive clinical trials while preserving safety

Drs. Copping from Genentech and Zou from Stanford University recently published the results of a study called Trial Pathfinder. Using the Electronic Health Records of 61,094 patients with advanced non-small-cell lung cancer, the team showed that data-driven eligibility criteria could lead to a more inclusive population.

Opposite to personalized medicine, restrictive inclusion criteria will not fully capture the efficacy and safety of a new drug. In one of its Guidance for the Industry Report, the Food and Drug Administration (FDA) encouraged the Pharmaceutical Industry to enhance the diversity of the Clinical Trial Populations.

Data-driven methodologies offer the potential for more inclusive trials without compromising safety or efficacy results.

Patient Adherence

Clinical trials can be cumbersome and difficult for patients. Once estimate that the dropouts rate among clinical trials is close to 30%. Patients are required to keep track of their medication intake and others criteria. This can lead to a lack of adherence, ultimately leading to unreliable results and dramatically extending the trial timeline and costs.

Wearable devices coupled with AI technologies offer the potential to develop real-time monitoring systems. By Picking up early warning signs leading to dropouts, AI and ML can help address individual adherence issues, which, when effectively addressed, will benefit the whole trial.

AICure is a startup that developed a suite of HIPAA and GDPR-compliant applications to improve patient engagement by using video between the patient and the investigatory site.

In Phase II clinical trial for Schizophrenia, the use of the AICure platform lead to a 25% increase in compliance.

Their AI platform also analyzes disparate data and combines them into actionable recommendations to optimize the research and business outcomes.

AliveCor develops wearable electrocardiograms devices. We reviewed the AliveCor KardiaMobile 6L and found it impressive. To make the most of the device, AliveCor recently partnered with Medable, which aims at providing decentralized clinical trials.

Such initiatives can improve recruitment, ease how data is collected, and ultimately improve patient adherence and clinical trials reliability.

What are the challenges of AI for clinical trials?

Despite the numerous potential benefits of Machine Learning, challenges exist that limit the reliability of the software. Among them, we can cite the lack of harmonization and the regulatory hurdles.

Garbage In, Garbage Out

Ai is only as good as the data ingested. One of the main hurdles is linked to the quality of the input data. Especially for Electronic Medical records (EMR), harmonization and interoperability are critical.

Apple is on a path to transform how research and individual care data are stored and processed. Apple ResearchKit allows gathering data through a customized framework. AI algorithms can make the most of these homogenized datasets to evaluate endpoints in clinical trials.

Beyond that, ResearchKit is also useful during Phase 4 (Pharmacovigilance) of the study, during which patients can easily report adverse events.

Wearable devices generate a huge amount of data that are often difficult to analyze. Through the Integrated Health Model Initiative, the American Medical Association aims at providing a framework allowing doctors and clinical trial investigators to make the most of the information collected.

Other big tech companies, such as Google with its Cloud Healthcare and Amazon HealthLake are also on a path to improve EHR interoperability and AI processing.

Regulatory bottle neck

According to the FDA regulations, AI software is considered a medical device. Therefore, it means that the AI software must be tested alongside the technology it substitutes.

Healthcare cybersecurity risks are a real threat. Regulatory hurdles are critical to ensure the privacy and protection of patient data.

What is the future of AI in Clinical Trials?

Machine Learning and Artificial Intelligence have tremendous potential for drug design and patient stratification and improving both end-points and adherence. However, adoption is still slow, and few major Contract Research Organizations (CRO) routinely implements such digital tools in their protocols.

Yet, the first necessary step is to homogenize the data and digitize the EHR effectively and reliably.

Was this article helpful?