Data Doesn’t Save Lives. Insight Does.

In the never-ending quest to improve—or at least tame—the byzantine task and exorbitant cost of drug development, the clinical trial is finally having its moment.

Often overlooked, the clinical trial phase has remained a largely manual, laborious process. Now, with the power of artificial intelligence (AI) opening a vast realm of real-world data (RWD) to researchers, we’re seeing progress and significant promise for improving clinical trials.

Already, researchers have effectively applied RWD to improve study design, optimize site selection, enhance patient recruitment, and improve patient safety.

But in the glow of this advance, it’s important to recognize the practical constraints on these developments. Volumes of data alone, even paired with enormous computing power, do not deliver medical breakthroughs.

Driving research advances with RWD requires high-quality, actionable data. To gain evidentiary insight from the new world of RWD requires the intimate oversight and involvement of collaborative teams of medical and technical experts.

The Challenge of Clinical Trials

The high cost of prescription drugs in the US is a complex issue with multiple contributing factors. One of these is the high cost of research and development, which is hamstrung by clinical trials that drag on for years, cost billions of dollars, and have an unfortunate 7.9% success rate.

Innovation is often cited as the key to gaining an advantage in the increasingly competitive global marketplace and remains key to discovering new treatments for illnesses such as diabetes, cancer, and heart disease. Life science organizations have a long track record of adopting new technologies to improve results: high-throughput screening, continuous manufacturing, targeted therapies, and so on.

Yet with the significant money and effort invested in the drug discovery process, there has been relatively little new technology to significantly improve the latter stages. Languishing in outmoded techniques, clinical trials have contributed to the rising costs and disappointing returns of drug development.

For a research phase in need of disruption, the promise of an RWD-fueled productivity leap is a welcome development for clinical trials.

Opening Doors With Real-World Data

Historically, the massive volume of health care data generated outside of traditional clinical trials has been underutilized. Much of this so-called RWD, for example, clinician notes in electronic health records (EHRs) and imaging reports, is unstructured, and not easily searchable or analyzable.

The observations, assessments, and treatment records within clinical notes offer a deep, invaluable understanding of the patient experience. But the difficulty in efficiently deriving validated meaning from this resource has limited the utility of EHR data in research.

Recently, the advent of AI and advanced machine learning (ML) and natural language processing (NLP) techniques have made it possible to curate and harmonize unstructured RWD, unlocking significant potential.

Putting RWD in Practice

The health care industry is witnessing growing adoption of AI-driven insights from unstructured RWD, including US Food and Drug Administration (FDA) guidance and a growing range of use cases.

When effectively deployed, RWD can be used to evaluate trial-eligibility criteria, recruit potential research participants and streamline recruitment. It can increase study efficiency, shorten timelines and improve patient access to research.

Data-driven trials informed by RWD start with a stronger foundation, potentially avoiding mismatched enrollment, unexpected side effects and costly delays that plague traditional trials.

Reflecting this potential, the Clinical Trials Transformation Initiative (CTTI) called for the modernization of clinical trials by “harnessing data from electronic health records, claims, registries, and other sources of real-world data to create new opportunities that can transform research, drive care, and achieve a fully integrated health process.”

Harnessing RWD in practice, however, is not a simple matter. To do so at scale requires skillful coordination and a sophisticated approach.

Not Just More Data; High-Quality Data

Expanding researchers’ access to data would seem to suggest an increase in opportunities to find meaningful clues and insight.

To that end, growing computing power and cloud-storage capabilities combined with advances in AI have encouraged some research organizations to assemble massive “data lakes,” with aspirations of filtering these vast resources for useful information.

But it’s important to note that successfully applying RWD to clinical research is predicated on much more than volumes of data; it requires access to rich, high-quality, curated datasets.

The outcome of any analysis will depend on the validity of the underlying data. For researchers, disease-indication specific datasets for therapeutic areas are essential. This means targeted information sourced from a variety of health care settings and data sources, and de-identified to protect privacy. With it, researchers can gain understanding of real-life disease progression, treatment patterns in diverse patient populations, and results over extended periods to aid in clinical decision-making and drug development.

Curating RWD for Insight and Evidence

Generating real-world evidence (RWE) from unstructured data entails a deliberate, clinically informed approach at each step—from ingestion to curation to application.

Teams of clinicians, nurses, clinical informaticians, data scientists, epidemiologists, biostatisticians, and engineers must work together to curate and standardize data while retaining its original clinical context.

To ensure high-quality, clinically validated data, processes must include the use of distinct training data, data harmonization, and clinician-led validation of any AI outputs. Models must be continuously refined to prevent bias and maintain accuracy.

Done correctly, these processes can provide researchers with access to high-quality, curated datasets for therapeutic areas that are disease-indication specific.

The Data Path to Better Outcomes

RWD is showing great promise and potential in clinical research. Analyzing de-identified patient pools can offer insights about efficacy, safety, and cost-effectiveness of a medical therapy beyond those obtained during traditional clinical trials.

In a drug development process commonly mired in cost overruns, delays, and disappointing results, AI-driven RWD analysis is a welcome evolution, with the potential to ease costs and accelerate timelines for bringing new and improved treatments to patients.

It is not, however, a panacea and there are no short cuts to results.

Data alone cannot transform pharmaceutical development and improve lives. Insight can only be generated from valid, meaningful data. Assembling and deploying RWD on a useful scale requires an effort that shouldn’t be minimized.

In today’s ultra-competitive global marketplace, innovation is essential. As research organizations race their competition—and outmoded research techniques—AI-curated RWD and the insights derived from it are poised to have a significant impact on the modernization of clinical trials.