The P95 Julius Clinical team recently joined researchers, regulators, industry experts and the wider pharmacoepidemiology community in Milan for the 42nd Annual Meeting of the International Society for Pharmacoepidemiology (ISPE).
Across the meeting, one message stood out: pharmacoepidemiology is playing an increasingly important role in how researchers, regulators and life science companies answer complex questions about medicines beyond traditional clinical trial settings.
Real-world evidence, artificial intelligence, causal inference methods, environmental health, and representation featured prominently across the scientific program. Together, these topics pointed to a future where access to more data creates new possibilities, while placing greater responsibility on researchers to ask the right questions, select suitable methods, and understand the populations represented in those data.
Here are five themes our team took away from ISPE 2026.
Real-world evidence is becoming increasingly relevant to regulatory decisions
The role of real-world data and real-world evidence in regulatory decision-making was a recurring theme at ISPE 2026.
The opening program included perspectives from the European Medicines Agency, while later sessions examined regulatory experience with RWE, including lessons from recent FDA approvals and the Advancing Real-World Evidence Program.
The discussion reflected both the potential and the practical challenges of using routinely collected health data to support regulatory questions.
Large healthcare databases, registries, surveillance systems, and other real-world sources can provide information that may be difficult to obtain through clinical trials alone. They can help researchers study larger or more varied populations, examine longer follow-up periods, and investigate how medicines perform in routine care.
Despite the huge amount of health data generated each year ~1 zettabyte (1 billion terabytes), much of the data remains unused or inaccessible for evidence generation. Also, despite the ~1 zettabyte (1 billion terabytes) of health data generated each year, much of it remains unused in RWE.
Data provenance, completeness, endpoint definitions, confounding, missing information, and the suitability of a data source for a particular research question remain central considerations. Regulatory use also requires transparency in study design, analysis, and assumptions.
For pharmaceutical and biotech sponsors, this means RWE planning increasingly needs to begin with the decision the evidence is intended to support. The research question, study design, data source, and analytical plan need to work together from the outset.
This is particularly relevant for post-authorization research, safety studies, effectiveness studies and evidence programs that follow a medicine across its lifecycle.
Target trial emulation is strengthening causal questions in observational research
Methodological discussions at ISPE reinforced continued interest in target trial emulation as a framework for causal inference using observational data.
The meeting included a dedicated course on target trial emulation, alongside multiple studies applying the framework across different therapeutic questions.
Clone-censor-weighting received particular attention. The method can help researchers study treatment strategies that change over time while reducing biases such as immortal time bias.
These methods matter because causal inference can be biased by immortal time bias, where observation time in subjects who receive an intervention includes a period during which they must remain event-free to receive the intervention, artificially inflating survival time and potentially overestimating the intervention’s effectiveness.
Researchers are often interested in what might have happened if one group of patients had followed one treatment strategy, and another group had followed a different strategy. Routine healthcare data were rarely collected with that comparison in mind.
Target trial emulation asks researchers to define an observational study in relation to the clinical trial they would ideally conduct. Eligibility criteria, treatment strategies, assignment, follow-up, outcomes, and analysis are specified before the observational comparison is made.
This approach helps researchers design robust observational studies that can be used to make explicit causal inferences and can expose sources of bias and make study assumptions clearer.
For sponsors, the message from ISPE was not that one methodology will solve every observational research question. Instead, increasingly sophisticated data require equally careful study design.
The quality of an RWE study still begins before the analysis.
AI is entering pharmacoepidemiology, but scientific oversight remains central
Artificial intelligence appeared throughout the ISPE program, from AI-generated target trial protocols to large language models for research summaries and AI-assisted clinical codelists.
The tone of these sessions was appropriately curious and critical.
AI has clear potential to support parts of research that require large amounts of information processing. It may assist with literature review, protocol development, coding, data preparation, and other research tasks.
At the same time, several sessions examined where AI-generated outputs can be incomplete, inconsistent, or incorrect. Despite significant improvements in AI performance and a reduction in hallucinated outputs, human-in-the-loop review remains a cornerstone of responsible AI use, ensuring that results are scientifically valid, contextually appropriate, and trustworthy.
The value of AI in research will depend on more than speed. Researchers need to understand how an output was produced, whether it can be reproduced, where bias may enter, and when expert review is required.
Drug discovery provides another reason to watch this field closely. One figure discussed at the meeting was that more than 175 AI-derived molecules are currently moving through R&D pipelines. Attention will increasingly turn from whether AI can identify candidates to whether those candidates produce convincing clinical results.
For pharmacoepidemiology, the same principle applies. AI can support research, but scientific judgment remains central when defining questions, selecting data, choosing methods, and assessing findings.
The most useful applications may therefore be those that combine computational capability with clear human oversight.
Environmental health is becoming part of the pharmacoepidemiology conversation
Another important theme in Milan was the relationship between medicines, healthcare, and the environment.
ISPE included a dedicated pre-conference course on environmental pharmacoepidemiology, examining methods and applications at the intersection of medicines, populations, and environmental exposure.
This reflects a broader question for health research: how should environmental factors be considered when studying health outcomes and healthcare interventions?
Climate-related exposures can affect disease patterns, vulnerable populations, healthcare access, and medicine use. Pharmaceutical products and healthcare systems also have their own environmental footprints.
For epidemiologists, this creates new research questions and new data requirements.
Environmental information may need to be linked with health records or other population-level data. Geographic variation, time-dependent exposure, and socioeconomic factors can introduce additional sources of confounding. Researchers may also need expertise that crosses traditional disciplinary boundaries.
The discussions at ISPE suggest that environmental considerations will become an increasingly relevant part of population health research and evidence planning.
Real-world data need to represent real-world populations
Perhaps one of the most important themes from the meeting was representation.
A dataset can contain millions of records and still provide an incomplete picture of the population affected by a disease or treatment.
Older adults, people with multiple health conditions, underserved communities and populations that have historically been underrepresented in research may differ from the groups most commonly studied in clinical development.
ISPE sessions on geriatric pharmacoepidemiology, health equity and data fitness brought this issue into focus.
The implication for RWE research is clear: size should never be confused with representativeness.
Researchers need to understand who is present in a data source, who is absent, and how those differences could affect the findings.
That includes examining age, sex, geography, socioeconomic factors, access to healthcare, comorbidities and other characteristics relevant to the research question.
For global studies, these questions have become even more important. Healthcare systems, diagnostic practices, data infrastructure, and access to treatment differ between countries and regions.
Working with local experts and understanding local data systems can therefore be just as important as selecting the statistical method.
What these themes mean for future real-world evidence programs
ISPE 2026 reinforced a central idea for pharmacoepidemiology: more data create more possibilities, but credible evidence still depends on careful scientific choices.
For sponsors planning RWE and epidemiology programs, several priorities stand out:
- Define the decision the evidence needs to inform before selecting the data source.
- Assess whether the available population represents the people relevant to the research question.
- Select causal inference methods according to the treatment strategy and potential sources of bias.
- Build transparency and reproducibility into analytical work.
- Apply appropriate scientific oversight when AI is used within the research process.
- Consider geographic, environmental, and health-system factors when planning multi-country studies.
- Plan regulatory evidence needs across the medicine lifecycle rather than treating RWE as a separate activity after clinical development.
These questions are increasingly connected. Regulatory evidence, methodology, technology, and representation cannot be considered independently if research is expected to reflect clinical practice.
They also point to a broader connection between clinical development and evidence generated after or alongside traditional trials. Questions raised during development can influence later RWE plans, while findings from routine care can add context to what is learned in controlled research settings.
For sponsors, bringing these perspectives together earlier can support stronger evidence planning across the product lifecycle.
Looking ahead
The conversations in Milan showed a field asking increasingly ambitious questions of real-world data.
AI can support researchers working with information at a greater scale. Target trial emulation can strengthen causal thinking. New sources of data can broaden our understanding of medicine use and outcomes. Environmental and health equity research can help researchers identify parts of population health that traditional datasets may miss.
Yet the foundations of credible pharmacoepidemiology remain familiar: a clear research question, suitable data, appropriate methods, transparency, and scientific expertise.
For P95 Julius Clinical, these themes connect directly with our work across clinical development, epidemiology and real-world evidence. By bringing clinical research experience together with epidemiological, statistical and data expertise, our teams support evidence generation across the development lifecycle, from study design and feasibility through post-authorization research, analysis and scientific communication.
Our experience across clinical trials and epidemiological research also gives us a broader view of the questions sponsors face as evidence needs develop over time. Clinical and real-world data answer different questions, but together they can provide a richer understanding of how medicines and vaccines perform across research settings and routine care.
As RWE plays a greater role in regulatory and healthcare decisions, the connection between clinical development and evidence generated from routine care will become increasingly important. Scientific quality, local understanding, and thoughtful study design will remain central to producing evidence that regulators, sponsors, and public health decision-makers can trust.
Interested in discussing your epidemiology or real-world evidence program? Connect with our real-world evidence experts today
