PHUSE US Connect 2026: AI and the Human Factor
I had a great time in Austin at PHUSE US Connect 2026, March 23-26. It was wonderful reconnecting with long-time colleagues, meeting new people, and seeing how the clinical data science ecosystem continues to evolve.
A few themes stood out across the conference.
1. AI is reshaping clinical data workflows.
Not surprisingly, AI was everywhere. Sessions explored how large language models and agent-based systems are being applied to statistical programming, automated dataset generation (SDTM/ADaM), and orchestration of end-to-end analytics workflows.
2. Automation through metadata and standards.
A second major theme was the automation of the clinical data lifecycle. Many talks focused on metadata-driven pipelines that move from protocol to analysis outputs with far less manual intervention.
3. Modernizing data standards and interoperability.
There was significant discussion around CDISC modernization, biomedical concepts, and initiatives like CDISC 360i, as well as the growing importance of interoperability with standards such as FHIR.
4. Real-World Evidence continues to gain momentum.
Real-world data and evidence had its own dedicated stream this year. Sessions addressed RWD quality, external control arms, and integration with clinical trial data.
I was also pleased that Orizaba Solutions presented work in this area focused on standards for submitting data derived from real-world sources to regulators. I’ll share that presentation soon.
Two keynotes that emphasized the human side of data
Despite the focus on AI, the keynote speakers highlighted something equally important: how humans interpret and communicate data.
Lynne Peeples opened the conference with a talk titled “When Data Doesn’t Speak for Itself.” She reminded us that biomedical data is often assumed to be objective and self-explanatory. The COVID-19 pandemic showed how easily data can be misunderstood or mistrusted when uncertainty and context are not communicated clearly.
She introduced a helpful framework for communicating data using the acronym CLEAR:
Context – Provide the background needed to interpret the data
Limitations – Be transparent about what the data cannot show
Explain – Clarify who collected and analyzed the data
Audience – Meet people where they are
Representation – Be thoughtful about how the data is presented
It’s a powerful reminder that good data science also requires good communication.
The closing keynote from Dr. Lilliam Rosario (TransCelerate) focused on how data standards enable collaboration and better regulatory decision-making.
Her message resonated strongly:
Standards give us a shared language that allows the ecosystem to learn together.
Structured clinical data allows knowledge to accumulate across programs, improving the fidelity of risk-benefit decisions and enabling new insights. Initiatives in toxicology data sharing and clinical trial data sharing are already demonstrating how harmonized data can unlock cross-study learning.
One idea she emphasized stuck with me:
Digitalization is not modernization. It’s risk mitigation.
When sponsors, solution providers, and regulators align on structured data exchange, safety insights can move through the system rather than being reconstructed again and again.
Final reflection
If there was one underlying message from PHUSE this year, it’s this:
AI may accelerate analysis, but structured data and shared standards are what make learning possible.
As the industry continues to expand the use of real-world data and AI-driven analytics, ensuring that data is well-structured, traceable, and fit for regulatory use will only become more important.
This is exactly the type of challenge we enjoy working on at Orizaba Solutions.
Looking forward to continuing these conversations with colleagues across the community.