Transforming Clinical Trials with AI

Clinical trials are considered the gold standard for assessing the safety and efficacy of novel drugs in development. An investigational drug typically progresses through a series of trials of increasing size, ultimately reaching a large trial comprising hundreds of patients. But despite their essential role in the biotech, medical device, and pharmaceutical industries, clinical trials are plagued with challenges that impede medical progress.

Developing prescription drugs is a high-cost, high-risk endeavor. Average research and development for an approved prescription drug requires an investment of $2.9 billion and takes more than 11 years. Clinical trials alone can cost an average of $1.1 billion over 6.6 years. In fact, clinical trials account for a staggering 40 percent of the pharmaceutical industry’s research budget. To make matters worse, only 14 percent of drugs that enter clinical trials are eventually approved.

The effectiveness of clinical trials in drug development is directly dependent on the quality of the data generated. Ideally, a trial would continuously collect objective data about a patient’s response to a drug. But this is often not practical using traditional approaches. Many diseases are monitored sporadically and rely on subjective physician ratings of disease status. For instance, diseases like multiple sclerosis and Parkinson’s disease are scored by clinicians roughly two to five times per year, despite the fact that a patient’s disease state fluctuates daily or even hourly. To address this limitation, approaches like hourly diary reporting by patients have been used, but these can be a significant burden.

Emerging hardware and machine learning (ML) technologies hold promise for tackling these challenges. Remote monitoring using wearables and other sensor-equipped devices now has the potential to regularly track patients, in their native setting, without substantial burden or bias. By collecting data like patients’ movements, heart rate, and glucose levels, these devices can help produce consistent, objective evidence of the actual disease state and the treatment’s impact. Not only can remote monitoring have cost benefits, it could also detect efficacy and safety issues conventional methods may miss.

Given the expected level of sensor adoption, new technologies for collecting and effectively using data are relevant and timely. In a study conducted for Intel, Kaiser Associates projected that up to 70 percent of clinical trials will incorporate sensors by 2025. To gain the most value from these devices, pharmaceutical companies must capture, manage, and analyze vast amounts of data. For example, a typical 100-person trial that runs for six months can generate over 200 billion data points. The Intel® Pharma Analytics Platform (Intel® PAP) addresses these issues, offering a holistic, device-agnostic platform, with data gathering, cloud, and artificial intelligence (AI) capabilities designed for a new era of clinical trials.1

The Intel® Pharma Analytics Platform (Intel® PAP)

Intel has developed a solution that captures new kinds of data from clinical trial subjects using sensors, wearables, smartphone apps, and other devices. The Intel® Pharma Analytics Platform (Intel® PAP) uses remote monitoring to passively and efficiently capture continuous clinical data from sensor-equipped wearable devices. This sensorial data, as well as electronic diaries and patient-reported outcomes (PROs), is transmitted (de-identified) to a secure cloud for storage and analysis—using ML and other AI methods to analyze the data. Collecting data from multiple sensors, the platform helps develop objective measures for assessing symptoms and quantifying the impact of therapies, and provide a broader understanding of patient health across a wide spectrum of needs.

Intel IT teamed up with the Michael J. Fox Foundation (MJFF) to improve research and treatment associated with Parkinson’s disease. The collaboration included a multiphase research study to gain insights from patient data collected with mobile technology, including wearables. The study results helped build the Intel® Pharma Analytics Platform (Intel® PAP). Since then, working with leading pharmaceutical companies, medical centers, research institutions, and CROs, the Intel® Pharma Analytics Platform (Intel® PAP) has been used in dozens of trials, comprising more than 1.5 million hours of data collection with over 1,000 patients.

This edge-to-cloud AI solution allows clinical teams to measure and collect patient data, including skin temperature, sweat levels, heart rate, blood pressure, glucose, and movement activity, both during the day and while sleeping. Data can be collected continuously as patients go about their daily lives, with no interaction needed. A mobile application allows data collection through questionnaires, digital diaries, and home assessment tasks. Patients can also get feedback on certain symptoms and manage their medication intake. Game-like elements help patients stay engaged and motivated.

This solution lets pharmaceutical companies and contract research organizations (CROs) collect data more efficiently. Trial administrators and CROs can track adherence in real time, intervening to encourage compliance with treatment protocols. Clinical teams can monitor patients for adverse events and intervene to help improve care and reduce the number of dropouts.

The platform has a rich ML library, with tools for researchers and data scientists to make queries and run algorithms on the data collected, securely and at scale. Intel also offers an analytics service for data analysis and clinical endpoint development by Intel’s ML specialists.

These capabilities make for more efficient and productive trials, which can get the latest treatments to patients sooner and more cost-effectively. The platform’s benefits are significant.

Greater accuracy. With automatic collection of consistent, unbiased, high-quality sensor data, pharmaceutical companies can measure changes more accurately. Researchers can use this objective evidence to better assess and demonstrate a new treatment’s clinical efficacy, safety, and side effects.

Deeper insights. Remote monitoring data produces a powerful foundation for analysis that can deepen insights into how a drug affects symptom progression and quality of life. Analytics teams can combine anonymized trial data with information from genomic, lifestyle, and other sources to explore new possibilities for future treatment breakthroughs. The platform was developed by data science specialists who have expertise in signal processing and ML, as well as experience developing clinical endpoints and objective measurements.

Increased patient retention. Remote data collection can help reduce the frequency of clinic visits and simplify the tasks clinicians and patients must perform. Reducing the burden on patients helps increase recruitment, compliance, and retention. Real-time communication can also improve compliance and help facilitate the industry’s movement toward virtual clinical trials.

Improved patient engagement. The solution delivers practical value to patients, which can improve loyalty that extends beyond the trial itself. Patients may gain a clearer picture of their health, an increased sense of control, and a feeling of pride in contributing to important research. Advocacy groups, patients, and families may be more loyal to drug companies that demonstrate their commitment to advanced research and innovative technologies.

Accelerated time to market. By producing high-quality data and reducing the dropout rate, trial leaders may be able to conduct shorter trials with fewer enrolled participants. With advanced analytics and larger, more diverse data sources, analysts can generate more robust evidence for regulatory agencies. Pharmaceutical innovators may also solidify their scientific understanding earlier in the development cycle. According to the Kaiser Associates study, this can lead to more efficient, evidence-based allocation of resources.

Built on a Proven Technology Foundation

The Intel® Pharma Analytics Platform (Intel® PAP) is scalable, allowing researchers to collect, store, and process data from thousands of patients. Once data is transmitted to the cloud, scientists and researchers can perform complex calculations and queries, and generate reports to gain new insights. While the initial focus is on neuromuscular diseases, the platform can be used for other diseases, types of studies, and different data sources.

The solution runs on cloud-based infrastructure at Amazon Web Services*. It takes advantage of the latest generation of Intel® Xeon® processors and Intel® Solid State Drives (Intel® SSDs) that deliver high performance for large data volumes. Intel® SSDs also allow for enhancing privacy and security through full-disk encryption.

Fulfilling the Hope of Clinical Trials

Building on Intel’s AI and analytics experience, the Intel® Pharma Analytics Platform (Intel® PAP) enables continuous collection and analysis of objective data from patients enrolled in clinical trials, with less burdensome methods and more accurate results. By augmenting subjective PROs with sensor data collection and applying ML and other AI methodologies, pharmaceutical leaders can capitalize on cutting-edge technology innovation designed to help reduce trial costs, improve data quality, increase patient compliance and engagement, and ultimately identify treatment efficacy. These changes may also help identify promising new approaches and accelerate time to market, bringing innovative treatments to patients faster.

To learn more about this solution and where it is being used, read the Intel brief Transforming Clinical Trials with the Power of AI.

Product and Performance Information

1 Kaiser Associates, Enterprise-purchased Wearables and Consumer Health Data Platform. May 2016. Confidential research conducted for Intel.