Use of Artificial Intelligence in Clinical Trials
I’d like this blog to convey two main points:
#1: There are many challenges today in the healthcare industry. The cost of healthcare is increasing, along with the number of people in need of treatment. There is a need to diagnose and treat conditions more accurately and quickly, and accordingly, there’s an increasingly urgent need to develop the tools to do that better and faster.
#2: Artificial intelligence (AI) can help solve many of the challenges described in #1.
This is the second in a series of blogs about the use of AI in the healthcare industry. In Part I, we took a high-level look at AI, including a few “everyday” and non-industry-specific uses, general AI concepts, AI-specific terminology, and overall benefits and concerns about the use of AI.
Here in Part II, we’ll look at the use of AI in processes related to clinical trials.
Clinical Trials: Many Processes, Many Opportunities
The path to discovering a new drug includes many processes, from discovering the molecule, to non-clinical and clinical studies, to manufacturing and marketing, with validation all along the way. Even within the realm of clinical trials, there are many subprocesses, and within a company, there are frequently many teams involved in those processes.
In general, if a process needs to be performed, automation can be incorporated to make the process faster and more accurate. Processes that involve repetitive steps are frequently good targets for AI, and with an increasing amount of data becoming available, AI and “big data” together can be utilized to perform work faster, sometimes with results that would have previously been difficult or even impossible to achieve.
Let’s look at a few processes in which AI can be used in the conduct of clinical trials:
- Clinical trial design, management and analysis
- Subject recruitment and selection
- Integration with Wearable Technology
- Randomization and Trial Supply Management (RTSM) systems
To fully understand the benefits of AI in clinical trials, it may help to remember a potentially overlooked challenge that I mentioned earlier: Due to global population increases and other factors, there is an increasing shortage of personnel in health care, including those who participate in the conduct of clinical trials. The use of AI in clinical trials has the potential to mitigate some of these shortages. According to a survey conducted by the World Health Organization (WHO) and Global Health Workforce Alliance (GHWA), there are an estimated 27.2 million skilled health professionals for a global population of approximately 6.7 billion people. By 2035—just 15 years away—an additional 1.9 billion people will require healthcare services. The survey also determined that 118 countries fell below an established threshold of 59.4 skilled health professionals per 10,000 population.i
To share one specific example from author-journalist Thomas Friedman’s book “Thank You for Being Late: An Optimist’s Guide to Thriving in the Age of Accelerations,” the population of Niger in 1950 was 2.5 million. Its population today is 19 million, and the United Nations projects that even with declining growth rates, the country’s population will be 72 million by 2050. That’s nearly a 30-fold population increase in 100 years.ii
Fortunately, the same WHO-GHWA survey concludes: “Despite all difficulties, the commitment of the international community to improving the health of all is stronger than ever.”iii I’ve seen that commitment in our industry over the years, and I’m sure that you have, too. Although the WHO-GHWA figures are for all healthcare—not just numbers involving clinical trials—the figures provide evidence for the need to meet the challenges in our industry.
Some of the world’s big players that did much to launch the current technology surge a few years ago now have specialty groups in the healthcare AI area, including Intel, Amazon (primarily through its widely used Amazon Web Services), Oracle (now via its Health Sciences entity), and Microsoft.
Clinical trial design, management, and analysis
A key characteristic of AI is that it is able to utilize large amounts of data. In fact, big data is really a requirement for AI: the more data, the better. Using AI, clinical trials teams have the advantage of identifying the best populations to participate in trials. Experts say that, over time, the use of machine learning will result in smaller numbers of subjects being needed on trials. The use of AI in this area has the potential to reduce repetitive steps, data entry, and quality control errors, resulting in better, faster, and more accurate gathering and analysis of data, leading to a better prediction of results.
Subject recruitment and selection
As with all components of a clinical trial, the subject recruitment and selection process requires time and costs money. Clinical trial teams are constantly in search of better and faster ways to recruit and select subjects for trials. Using AI, study personnel have the potential to better identify patients who would be good candidates, as well as those who are likely not to participate. Over time, AI is expected to be able to identify candidates more quickly.
Wearable devices can provide several advantages, such as the ability to record real-time data that can provide more accurate data, increase adherence of subjects to protocol, increase convenience for subjects, and reduce operating costs. Whenever a device can directly and automatically record data such as heart rate or body temperature, it helps to ensure that the data is recorded accurately and whenever it is required, while reducing the likelihood of human transcription errors. In most cases, wearable devices record the information and transmit it to a smartphone, where it’s transmitted to the applicable research team, pharma company, or clinician, where the data is maintained for processing. Smartphone apps can also be used to record subject diary information. An additional benefit is that quicker and more specific feedback can result in greater subject engagement, further aiding in the recruitment, selection, and retention of subjects.
Pharmacovigilance (PV) involves detection, evaluation, and prevention of adverse events (AEs) and other drug-related problems.
According to research published in Pharmaceutical Medicine in 2019, “The volume of individual case safety reports (ICSRs) increases yearly, but it is estimated that more than 90% of AEs go unreported. In this landscape, embracing assistive technologies at scale becomes necessary to obtain a higher yield of AEs, to maintain compliance, and transform the PV professional work life.”iv
As with clinical trial data, PV processes frequently require repetitive steps and can be good candidates for the use of AI. AI is used in the context of existing PV processes. From the point at which the adverse events case enters the workflow, the information is coded according to company standards and industry-standardized coding dictionary (MedDRA). Adjudication of relatedness needs to be performed (a potential effective use for AI), as well as the medical assessment of the seriousness of the event. The resulting information is then transmitted to regulatory authorities.
Randomization and Trial Supply Management (RTSM) and Interactive Response Technology (IRT) systems
RTSM/IRT systems are used for randomizing and assigning subjects into treatment groups and for managing supply, including product dispensing. For years, companies that develop and host RTSM/IRT products have been bringing advanced technology to their solutions, including hosting RTSM/IRT as a “Software as a Service” (SaaS). Interactive voice technology (and later, interactive web technology) have been used by subjects for years for product dispensing. A number of RTSM/IRT solutions can be integrated with the Electronic Data Capture (EDC) system used to manage the clinical trial data.
Some solutions use preestablished algorithms to automate processes and can communicate treatment assignments via a preferred medium, such as email, text messages, or on accredited websites. Dosages can be calculated based on patient information, and AI can make the process more reliable.
Considerations and Potential Concerns With AI
The ubiquitous presence of AI and globalization of technology was made possible by many factors coming together in a short period of time, including rapid advances in processor speed, storage capacity, data transmission speed, cloud technology, portable computing devices such as smartphones and tablets, and the resulting industry that now develops applications for those devices.
Although these factors have resulted in new and extraordinary possibilities for the healthcare industry, some risks may need to be addressed. In Thomas Friedman’s aforementioned book, he speaks of the changes resulting from the exponential accelerations in technology, and parallel increases in globalization. As Friedman says, “Globalization has always been everything and its opposite.”v
Some areas to be addressed include:
- Lack of knowledge. As mentioned in Part I of this series, some companies have expressed a lack of knowledge or understanding about AI. This may be especially true in smaller companies where there are fewer resources to research potential uses of AI.
- Lack of need. Some firms may sense that they have a lack of volume or need for AI and other automated processes.
- Cost of implementation and validation. Although there may be initial costs of time and money necessary to implement AI technologies, the benefits should outweigh those costs over time. In addition, many solutions are now offered to clients with many of the validation activities already completed by the supplier.
- Data Volume. Clinical studies and their study data, study monitoring reports, and pharmacovigilance information sets continue to generate increasing amounts of data. According to Kaiser Associates, a typical Phase 2 clinical trial with a six-month duration and 100 subjects would likely generate over 200 billion data points.vi However, storage media, including media hosted by solution providers, should now be able to handle those larger amounts of data, as well as backup and recovery capabilities. This is an area that should be confirmed by potential customers during due diligence.
- Security. In addition to ensuring that proprietary information is kept secure, it’s important to ensure the privacy of everyone involved in the trials, including study subjects, clients, and clinicians. Fortunately, in the continually developing world of IT security, protocols can now help to ensure encryption of data at rest on servers as well as during transmission. However, it’s important for subject matter experts to confirm adequate security.
When I started in the healthcare industry more than 30 years ago, it was common for healthcare companies to have IT departments that partnered with their respective business areas to develop numerous internal applications for performing business needs. Today, there are many more suppliers for those solutions, and the use of cloud technology has made it possible for suppliers to develop and host applications and perform much of the validation, saving time and money for pharmaceutical companies, CROs, device manufacturers, and other stakeholders. In turn, these advances gave rise to a platform greater than the sum of its parts, with the next phase, artificial intelligence, providing opportunities across many processes in the realm of clinical trials.
In summary, through the faster processing speeds and capacity, cloud technology, and other advanced technology, AI will continue to provide results that would have been difficult or impossible to achieve otherwise just a few years ago. In fact, the challenges mentioned at the beginning of this blog have truly become opportunities.
i. “A Universal Truth: No Health Without a Workforce.” World Health Organization. 2014.
ii. Friedman, Thomas L. Thank You for Being Late: An Optimist’s Guide to Thriving in the Age of Accelerations. Farrar, Straus and Giroux, New York. 2016. p. 175.
iii. World Health Organization.
iv. Mockute, Ruda et al. “Artificial Intelligence Within Pharmacovigilance: A Means to Identify Cognitive Services and the Framework for Their Validation.” Pharmaceutical Medicine, 2019.
v. Friedman, p. 154.
vi. Admati, Chen; Dolan, Yonatan; and McManus, Michael J., PhD; “AI and Wearables Bring New Data and Analytics to Clinical Trials,” Intel Solution Brief, Intel Corporation. 2017.
Check out Part I of this series here. For questions and inquiries, please visit our Solutions homepage or contact email@example.com.