AI speeds trials but patient trust and data quality still

Artificial intelligence and digital tools are transforming clinical trials, but their success depends on patient participation and the quality of data collected. Specialists emphasize that external control arms and digital twins can simplify research by cutting down on traditional control groups and modeling how patients might progress through treatments. However, these methods require precise, unbiased data to function effectively. Poor-quality inputs risk introducing errors and skewing results.

Aten explained that external control arms leverage real-world data to reduce the need for enrolling participants in control groups. Digital twins, which create virtual patient replicas, could model how enrolled patients might progress through a study, helping researchers account for differences among participants and improve statistical analysis. Yet both approaches share a critical vulnerability: the data they rely on must be free of systematic bias. “If the underlying data that you’re using still has systematic confounding and bias, that could be problematic,” he said, noting that randomization remains an important tool for reducing bias in research.

Regulatory bodies are adjusting to AI’s expanding role in drug development, though concerns about data consistency and traceability persist. Corso stressed the importance of transparent documentation for data collection, measurement, and source verification. Even routine measurements, such as blood pressure readings, can vary significantly in clinical settings. “There are a lot of ways to take a blood pressure in clinical research,” Corso said, emphasizing that measurement consistency matters when real-world data supports research findings.

Many trials continue to depend on antiquated workflows. Citing a Tufts study, Corso said 67 percent of informed consents in the United States are still completed on paper. Some consent documents can run dozens of pages, while sites often use isolated spreadsheets and manual processes, creating inefficiencies. “What we need is less paper,” Corso said. “And then systems that talk to each other.”

Patient engagement remains a significant obstacle. Franchetti, a nurse with extensive clinical research experience, argued that technology alone cannot improve trials without direct patient involvement. Patients should be consulted during trial design, data analysis, and AI implementation. Their apprehensions—such as fears of being replaced by automation—are increasing. “Am I going to be replaced?” Franchetti said, describing a question patients are asking as AI becomes more prominent in clinical research. Early collaboration can simplify procedures, align priorities, and reduce participant dropout rates. Currently, however, many decisions affecting trials are made without patient input.

Improving data infrastructure is a priority. While electronic health records have laid a foundation, gaps in accessibility and interoperability persist. Aten highlighted federal efforts, including the FDA’s initiatives and programs from the Office of the National Coordinator for Health Information Technology, as progress toward unified systems. The objective is to enable smooth data sharing while preserving public trust. Life sciences research, however, still uses multiple standards that do not always work together, unlike healthcare, which has adopted widely accepted protocols. These inconsistencies hinder collaboration and data exchange.