SPEAKERS:
• Maria Galindo Perez, VP, Global Head of Publications, Bayer
• Isabelle Lonjon-Domanec, VP, Clinical Medical Regulatory, Novo Nordisk
• Samantha Scheer, Head of Global Medical Communications, Neurology & Immunology, Merck KGaA
• Sarah Frazer, VP, Head of Medical Strategy, Real Chemistry
KEY TAKEAWAYS:
• Publication PDFs are largely invisible to LLMs; press releases dominate AI-mediated evidence retrieval
• Medical communications must enter the evidence lifecycle at gap identification, not manuscript submission
• Medical communications can act as a cross-functional integration point, connecting scientific, clinical, patient and field insights to shape more effective evidence generation and communication strategies
• Not only subgroups, but any data to be generated or literature gap to be filled without proper insight, will resolve only internal questions
• A Clinical Medical Manager role demonstrates that bridging trials and medical affairs requires dedicated mechanisms for earlier insight capture
• Medcomms measurement must evolve beyond reach metrics toward evidence of knowledge gain, decision-making impact and whether field questions progress over time.
When efforts were made to examine how evidence surfaces for patients and HCPs, the finding was blunt. “Many of the things we create, where we invest significant time, effort, and resources, are not visible,” said Maria Galindo Perez, VP, Global Head of Publications, Bayer, at Pharma 2026. This underscores the importance of understanding where scientific information appears and how evolving digital ecosystems influence what is surfaced and accessed.
Sarah Frazer, VP, Head of Medical Strategy at Real Chemistry, framed what this means operationally: "We are actually using AI not just to push out content more efficiently, but also to listen. We're scanning what healthcare providers are asking, what patients are searching for, and what signals are coming back almost in real time." The gap those signals reveal is not a technology problem. It is a structural one. The organizations that have begun closing it share a common insight: the visibility failure downstream is inseparable from a positioning failure upstream.
Rigor Created Formats AI Cannot Read
The LLM invisibility problem has a direct upstream cause. Medical communications has been positioned as the terminal output of the evidence pipeline, a content function that receives completed data and translates it for audiences. This sequencing guarantees the wrong artifacts get produced.
"We are still considering scientific communication as content and not as part of the evidence strategy," Galindo Perez argued. "It's part of this end-to-end from the day that you say, hey, there is a gap, we need to fill it. You create the data and translate it intentionally, then you bring it to the field and close the loop by bringing back insights." When translation is separated from generation, the data produced tends to answer the questions the generating team had, not the questions practitioners in the field are actually asking.
That is why medical communications needs to be positioned not only as a downstream content engine, but as a cross-functional integration point for insight. Its strategic value lies in connecting what clinical teams are generating, what medical affairs is hearing, what patients and HCPs are asking, and what field teams need to support better decision-making. When that integration happens early, evidence generation and communication strategy can be designed around real external needs rather than internal assumptions.
The concrete version of this failure is striking. "Take the case we have great data, phase 3, simultaneous publication, and once it's out, we already have planned 50 subgroup analyses without waiting a second, talking to the doctors and the patients and seeing what are the questions that they have," Galindo Perez noted, "because maybe our 50 subgroup analyses are not answering what they want to know." This is the predictable output of a structure where medcomms enters after the scientific agenda has been set. The resource allocation is substantial; the relevance to field decision-making is assumed rather than verified.
Samantha Scheer, Head of Global Medical Communications, Neurology & Immunology at Merck, identified where this structural problem surfaces next. Medcomms has achieved cross-functional collaboration, she argued, but the frontier is personalization: "getting the right data to the right audience at the right time through the right channel. Where I think we have the challenge now is making that more personalized." Targeting broad audience categories, HCPs, patients, nurses, is table stakes. The unsolved problem is educational need at the individual level, in a high-volume information environment.
Frazer connected that challenge directly to the AI environment: "Thinking about publications, for example, what journals are showing up; thinking about forums, for example, patient forums, we can look at them earlier and actually understand where the information needs to be put out." The compliance architecture that governs publication formats, gated PDFs, image-heavy infographics without indexable text, dense manuscripts written for peer reviewers, was designed for a different retrieval environment. The same rigor that ensures scientific validity is now producing formats that AI systems cannot parse. Digital opinion leaders, whose regulatory status varies by jurisdiction and whose scientific accuracy is uneven, operate on the exact channels, YouTube, social media, short-form digital, that AI systems index most aggressively. Pharma's highest-integrity evidence is structurally excluded from the discovery environments its audiences now use.
Three Approaches to Earlier Insight Integration
Each of the three companies represented on this panel described a different approach to bringing insight closer to evidence generation and communication planning. All three share one feature: they move medcomms authority, or insight flow into medcomms, earlier in the lifecycle. "Getting the insights at the right time can really impact the quality and speed of medical communications further down the road," Frazer noted, and each model addresses a different point at which timing fails.
Novo Nordisk's approach is a dedicated bridging role. Isabelle Lonjon-Domanec, VP, Clinical Medical Regulatory, described the Clinical Medical Manager: "That person is working closely with clinical operations, is not part of medical affairs, but has a link with the medical affairs community. Those people are really training the clinical operations from a scientific point of view, but having also a very in-depth discussion with investigators to collect those insights." The CMM operates during Phase 2 and 3 trials, capturing signals about how the drug is experienced in practice before launch planning begins. Sitting outside Medical Affairs, the role avoids the prioritization pressures that push MA engagement later, while maintaining a deliberate connection to the communications function.
The downstream payoff is concrete. Lonjon-Domanec described a liver disease drug where investigators flagged rash as a recurring side effect: "We established early on a publication committee to develop a publication on rash management among patients with liver disease, and we brought together hepatologists and dermatologists early on so that when the drug was on the market it was much easier for them to handle that side effect." Pre-launch publication activity, driven by trial-phase insight, meant HCPs had actionable guidance at the moment they needed it.
Bayer's model achieves upstream integration differently. Galindo Perez described consolidating all publication leads under a single centralized team. They continue to work closely with their respective therapeutic area strategy teams but sit organizationally together. When a communication challenge is resolved in one therapeutic area, the solution transfers immediately across others. Publication expertise pooled organizationally accelerates problem-solving and prevents each TA from discovering the same obstacles independently.
Merck's MS in the 21st Century initiative represents a third model: co-creation as an integration mechanism. "It is made up of both physicians and patients from the MS community," Scheer explained. "They focus very much on what are the true challenges, what are the real unmet needs, and therefore they're able to produce tools and resources that are actually really relevant to the community and can support other MS patients and caregivers." This approach moves the audience from recipient to co-designer, achieving early insight integration not through a bridging role or a centralized team, but through embedded stakeholder participation in the development process itself.
Discover more on this topic at Pharma Commercial Data & Tech Europe 2026 (4-5 November, London) Europe’s collaborative home for data and tech pioneers. Visit the website here.
Measuring What the Loop Is Doing
The measurement question is where candor is scarce. "Do we actually know if any of this is working?" Frazer asked directly. The honest answer from the panel: not well, and for structural reasons that mirror the upstream problem.
"For every deliverable, you need to be very clear up front what is the objective and then how are you going to measure success," Scheer argued. "I do think we need to push ourselves to think a little bit more qualitatively, did this change clinical practice, did this change behavior? That is more challenging to define, but it is the real measure." Citation counts and download metrics are easy to generate and nearly useless as evidence of impact. They measure production and reach, not whether clinical decision-making changed.
The broader implication is that medical communications measurement must move beyond metrics that indicate whether information was distributed or accessed, toward metrics that show whether it changed understanding, confidence or decisions. Reach still matters, but it is an incomplete proxy. The stronger question is whether the communication has improved knowledge, resolved uncertainty, supported clinical decision-making, or moved audiences to a more sophisticated stage of engagement with the evidence.
Galindo Perez offered a more operational proxy: watch what questions return from the field, and track whether they evolve. At launch, the field asks whether the drug is efficacious and what the safety profile looks like. Those questions should resolve. If they persist for months, the communication program is not working. The signal that it is working is that questions develop, from efficacy to patient selection, from safety to management, indicating that practitioners are progressing through the learning curve rather than circling the same knowledge gaps.
Lonjon-Domanec acknowledged the organizational reality plainly: "The word that comes to my mind is humility, because I have worked for several companies and we can always improve, we can always do better, but it's key that we gather those patient insights because ultimately we want to improve patient lives. But it's a journey." Patient insight integration is immature across the industry, and the tools to accelerate it, AI-assisted signal detection, cross-affiliate consolidation, patient advocacy co-construction, are early in their deployment.
The Full Loop Requires Redesigning the Evidence Artifact
The insight-integration approaches described in this session, bridging roles, centralized teams, co-creation models, address the upstream failure.
They do not yet solve the downstream visibility problem. That requires a further category of change: redesigning evidence artifacts for the retrieval environments that now dominate.
Galindo Perez described what this looks like in practice. Publication titles and abstracts need to be written to answer the questions HCPs and patients are actually asking, although there may currently be some misalignment with peer‑review conventions. Infographics require indexable text. Press releases, which LLMs already surface, need to carry accurate references so that AI-mediated patient searches lead back to primary evidence rather than dead ends.
"The more we put out there in the right channel the correct data, then the AI will use it and deliver when a patient or an HCP asks the question," Lonjon-Domanec observed. The logic is straightforward; the execution requires medcomms teams to think simultaneously about evidence validity and retrieval architecture, two disciplines that have not historically shared a planning table.
"Stop thinking of medical communication as content creation," Galindo Perez concluded. "It is this strategic piece of the end-to-end data generation, and be very intentional on how you translate that data and how you communicate that data." Scheer's directive for where to start is equally precise: "Start early to understand your audiences. Who are your key audiences? What are their educational needs? How do they like to consume content? Because this is foundational for the communication strategy going forward."
The organizations that have already built approaches for earlier insight integration have addressed the first half of the loop. The second half is an artifact problem. AI is not failing to surface pharma's evidence because AI is inadequate. It is surfacing what it can find. Companies that continue producing formats optimized for peer-review infrastructure, while their audiences search through LLMs, are not facing a communications gap. They are facing a strategic choice about whether evidence is designed to be found.
To get you highlights of Pharma 2026 faster, we are using generative AI technology to summarise the transcripts of the sessions. If you have any feedback about the summary, please contact lucy.fisher@thomsonreuters.com.
Discover more on this topic at Pharma Commercial Data & Tech Europe 2026 (4-5 November, London) Europe’s collaborative home for data and tech pioneers. Visit the website here.
