ISPOR 2026 daily round-ups: Insights from Day 3

Day 3 at ISPOR 2026, the annual meeting of ISPOR—the Professional Society for Health Economics and Outcomes Research, brought the conference to a close with a focus on global perspectives, emerging methodologies, and the future direction of HEOR.

The program closed with final plenary, “Innovation Under Pressure: How Will US Drug Policy Reshape Innovation, Evidence, and Access Globally?”, which examined how policy developments are influencing global markets, evidence generation, and long-term incentives for innovation. A spotlight session on transparent and validated AI workflows in HTA further highlighted the growing emphasis on credibility and trust in AI-enabled evidence.

Across concurrent sessions, discussions explored generative AI in HEOR workflows, patient-centered trial design and economic models, and the increasing role of global data sources, including China’s emergence as a hub for evidence generation. Here are some of the highlights from Day 3.

We hope our daily round-ups have provided useful insights from across the meeting. In the coming weeks, we will continue our coverage with deeper dives into key sessions, Peek Behind the Poster interviews, and post-event reflections.

Closing remarks and leadership updates

The final session of ISPOR 2026 began with Leadership Reports, reflecting on both ISPOR’s progress over the past year and its future direction, with a strong emphasis on strengthening the connection between rigorous HEOR science and real-world policy impact. Opening the session, ISPOR CEO Rob Abbott highlighted the conference’s focus on ensuring HEOR informs decision-making, pointing to discussions on US pharmaceutical policy, financial toxicity in cancer care, and the global consequences of healthcare policy decisions. Abbott emphasized that advancing a world where healthcare is “accessible, effective, efficient and affordable for all” requires not only generating high-quality science, but ensuring that evidence reaches policy and implementation settings where it can improve patient outcomes.

Providing the science update, Eberechukwu Onukwugha (ISPOR Past President) outlined a productive year for ISPOR’s scientific activities, including updated HEOR competency frameworks, four Good Practice Reports in progress for publication, and continued work on frontier topics such as AI, RWE, value-based healthcare, patient centricity and whole health. AI was highlighted as the leading HEOR trend for 2026, while new competencies around AI, equity, policy and communication have been incorporated into ISPOR’s guidance. Onukwugha also introduced a draft whole health definition, encouraging broader consideration of health outcomes beyond treatment alone, incorporating prevention, caregiver burden and wider social determinants of health. Joining the discussion, Kelly Saldana (ISPOR Institute for Healthcare Transformation) described the Institute’s role as a bridge between evidence generation and implementation, helping ensure robust HEOR science translates more effectively into healthcare decisions and real-world practice globally.

Closing Keynote

In the closing keynote at ISPOR 2026, John Singer (Blue Spoon Consulting) challenged attendees to rethink some of healthcare’s most deeply embedded assumptions, arguing that many of the pressures facing healthcare systems today stem not simply from policy change, AI, or market disruption, but from an obsolete framework underpinning how innovation itself is conceptualized. Framing healthcare as operating within an “antique narrative,” Singer suggested that longstanding priorities around cost containment and crisis management have dominated healthcare thinking for decades without successfully solving either challenge. Against a backdrop of Medicare price negotiation, discussions around international reference pricing, AI-enabled drug discovery, and rapidly evolving therapeutic innovation, he argued that healthcare may be attempting to solve modern problems using frameworks built for a very different era.

Singer proposed shifting focus from healthcare delivery alone toward what he described as the “production of health” – emphasizing cardiometabolic health, mental health, reproductive health, longevity, and prevention at population scale. He argued that future competitive advantage may increasingly depend not on producing scientific breakthroughs alone, but on building systems capable of absorbing, scaling, distributing, and implementing innovation effectively. Highlighting examples ranging from gene therapies to GLP-1 medicines and employer-led healthcare models, Singer suggested healthcare is entering an era requiring stronger collaboration across pharmaceutical companies, technology developers, providers, employers, and health systems.

Concluding his remarks, Singer challenged attendees to consider whether HEOR could play a broader role in shaping future healthcare systems – not simply measuring value within existing frameworks, but helping redefine how value itself is understood.

“Your evidence cannot land any harder than the frame allows,” he concluded, arguing that healthcare innovation may increasingly depend on developing new narratives alongside new technologies.

Plenary 3: Innovation Under Pressure: How Will US Drug Policy Reshape Innovation, Evidence, and Access Globally?

The third and final plenary session at ISPOR 2026 examined how evolving US pharmaceutical policy may reshape innovation, evidence generation, and access globally. Moderated by Indranil Bagchi (GSK US), the panel featured Lotte Steuten (Office of Health Economics), Darius Lakdawalla (University of Southern California), Jens Grueger (Boston Consulting Group), and Lizheng Shi (Tulane University School of Public Health and Tropical Medicine) discussing topics including MFN pricing proposals, international reference pricing, global launch strategies, pharmaceutical innovation, and the growing influence of US policy decisions on healthcare systems worldwide. Speakers highlighted concerns around uncertainty, affordability pressures, and maintaining incentives for innovation, while also exploring how increasingly interconnected healthcare systems are reshaping evidence generation, reimbursement approaches, and patient access globally. A recurring theme across the discussion was the importance of balancing affordability with long-term innovation, alongside the growing role of HEOR in helping policymakers navigate increasingly complex healthcare environments. Closing the session, Steuten called on the HEOR community to “protect the value of science,” emphasizing the need for robust evidence and thoughtful evaluation frameworks as healthcare systems adapt to a rapidly changing global policy landscape.

Read our in-depth summary of Plenary 3 here >>>

Spotlight Session: Beyond Black Boxes: Case Studies of Transparent, Validated LLM Workflows for Accelerating Global HTA Submissions and Decisions

Moderated by Beth Devine (CHOICE Institute, University of Washington), ISPOR President-Elect (2025–2026), the session examined how large language models (LLMs) are increasingly being integrated into HEOR and HTA workflows, while also highlighting the need for validation, transparency, and scientific oversight. Devine framed the discussion around the rapid evolution of AI, from rules-based systems to generative and “frontier” AI, emphasizing that established RWE fit-for-purpose principles including reliability, reproducibility, transparency, and validity “apply in the world of AI as well.”

Lockwood Taylor (Flatiron Health) discussed the use of LLMs to extract clinically rich unstructured electronic health record (EHR) data for evidence generation and HTA. He outlined the VALID framework, which incorporates variable-level performance metrics, automated verification checks, and benchmarking analyses to assess LLM-derived data quality. Using prostate cancer and pan-tumor case studies, Taylor demonstrated that LLM extraction could achieve performance comparable to human abstraction when appropriately validated and adapted for country-specific documentation practices. He noted that “the power of LLMs to extract EHR data at scale… is undeniable,” while stressing that rigorous quality assessment remains critical.

Tim Reason (Estima Scientific) focused on AI-assisted systematic literature reviews and network meta-analyses (NMAs), arguing that the tools required to support AI-enabled HTA workflows are already available. He described applications including automated screening, extraction of NMA-ready data, AI-supported model selection, and report generation. Reason argued there is a “moral imperative” to accelerate analyses because delays in evidence generation can postpone patient access to therapies. He also highlighted the potential for “living” NMAs and continuously updated HTA systems supported by retrieval-augmented generation workflows.

Bill Malcolm (Bristol Myers Squibb) provided an industry perspective on the operational burden associated with HTA submissions and how AI-enabled workflows are being used to reduce timelines for economic model adaptations and dossier preparation. Malcolm emphasized that validation must remain “mandatory and continuous,” outlining considerations including reproducibility, blinded benchmarking, and peer-reviewed publications of AI toolchains. He described internal case studies where AI-supported workflows reduced adaptation timelines from months to days and enabled more rapid submission preparation, including in lower-resource markets. Across the discussion, the speakers consistently emphasized that AI should augment rather than replace HEOR expertise, with subject matter experts remaining “at the helm” of evidence generation and quality control.

Concurrent sessions

Productivity Gains From Generative AI Across the HEOR Workflow: Successful Case Studies

This breakout session explored how generative AI is beginning to reshape HEOR, with speakers emphasizing that future gains will come not simply from working faster, but from redesigning evidence workflows while maintaining scientific rigor. Moderated by Uwe Siebert (UMIT TIROL – University for Health Sciences and Technology; Harvard Chan School of Public Health), the session featured perspectives from (Harvard Medical School / Massachusetts General Hospital), Ipek Ozer Stillman (Takeda) and Turgay Ayer (Value Analytics Labs). Siebert highlighted growing pressure on HEOR teams to deliver more evidence, across more markets and scenarios, within increasingly compressed timelines, while cautioning that trust, validation, and governance remain critical as AI becomes embedded into structured workflows.

Representing academia, Chhatwal presented case studies applying AI to health economic modeling, including model validation and replication exercises. One example demonstrated how AI-assisted validation identified structural issues, verified formulas, and conducted cross-validation exercises in around an hour – tasks that traditionally take weeks. However, Chhatwal stressed that expert oversight remains essential, particularly for interpretation, calibration, and methodological judgment.

From an industry perspective, Stillman discussed Takeda’s experience piloting AI tools across modeling, value dossiers, literature reviews, and evidence generation activities. She emphasized that productivity should be viewed strategically rather than simply operationally, arguing that time savings should be reinvested into improving evidence quality, expanding analyses, increasing methodological rigor, and enabling earlier decision making. Stillman also highlighted the importance of governance, validation frameworks, and building organizational confidence in AI adoption.

Representing the vendor perspective, Ayer described agentic AI approaches designed specifically for HEOR workflows, presenting examples where multi-agent systems generated complex multi-country landscape assessments with hundreds of references in under 48 hours.

Across the discussion, speakers agreed that AI is increasingly moving beyond experimentation and into practical implementation – but emphasized that human expertise, transparency, and scientific rigor must remain central as HEOR workflows evolve.

Patient-Centered Clinical Trial Design: Using Patient Experience Research to Improve the Patient-Centeredness of Patient-Reported Outcomes Measures and Patient Preference Research to Prioritize Patient-Centered Endpoints

Designing patient-centered clinical trial endpoints requires a stronger understanding of how patients experience symptoms, treatment impacts, and benefit-risk trade-offs, according to speakers in this session moderated by Kelly Myers (RTI Health Solutions).

Jason Roberts (FDA) discussed how patient preference information (PPI) can support medical device development and regulatory decision-making across the total product lifecycle. He explained that “what is important to patients is critical in device development activities,” particularly when selecting outcomes, defining target populations, and evaluating benefit-risk trade-offs. Roberts also stressed that evidence used in regulatory submissions “must constitute valid scientific evidence.” 

Holly Peay (Faegre Drinker Biddle & Reath) presented qualitative research exploring patient experiences of genitourinary syndrome of menopause (GSM), a heterogeneous and often under-recognized condition. The findings highlighted substantial variability in symptoms, stigma, and day-to-day impacts, including effects on relationships, social participation, and self-concept. Participants also described limited communication with clinicians, with one patient noting that “it would be really nice if my healthcare providers had been talking about this.” 

Reflecting on the challenges of evaluating existing patient-reported outcome (PRO) measures, Peay explained that “it was really hard from the qualitative to figure out what to do for the PROs because of the heterogeneity.”

Myers then presented findings from a quantitative best-worst scaling survey designed to prioritize meaningful aspects of health and identify patient-centered clinical trial endpoints. The results demonstrated substantial heterogeneity across patient subgroups, with urinary frequency and urinary urgency identified as the most bothersome symptoms. Myers explained that the aim is to support selection of endpoints “that are more accurately reflective of patient priorities.”

Rewriting the Global Evidence Playbook: China’s Emergence as a Strategic Hub for Evidence Generation

The session explored how real-world data (RWD) and real-world evidence (RWE) from China can support global pharmaceutical development, market access, and regulatory decisions.

Lung-I Cheng (AESARA) framed the discussion around three interconnected themes: supply, demand, and capability. He highlighted the growing role of Chinese biotech innovation in global R&D, increasing pressure for globally coordinated evidence strategies driven by pricing and launch sequencing dynamics, and the rapid maturation of China’s evidence infrastructure.

Sheng Feng (LinkDoc Technology) outlined China’s fast-evolving RWE infrastructure and policy environment. He described how regions including Hainan and the Greater Bay Area can use locally generated RWD alongside US and European clinical trial data to support regulatory approvals without traditional local randomized controlled trials. Feng also reviewed how provinces including Beijing, Fujian, and Shanghai are opening large-scale claims and EMR datasets for market access, pricing, and health economics research. Additional developments included AI-enabled data curation practices, new governance laws governing health data use, and the creation of a dedicated IRB in Beijing for AI and RWE studies. Feng also discussed China’s “RCT Duplicate” initiative, which seeks to replicate and stress-test randomized trial findings using insurance claims data, including in complex areas such as oncology.

Kasey Fu (Vertex) examined the challenge of external validity and argued that epidemiology can act as a “contextual bridge” between Chinese and non-Chinese populations. Using examples in lung cancer and rare disease forecasting, Fu highlighted how epidemiologic measures may help assess comparability in standards of care, screening patterns, and diagnosis practices across jurisdictions. She also noted that variability across China’s healthcare system may create opportunities to identify more comparable subpopulations for global studies.

Presenting slides on behalf of Hui Huang (CSD Partners LLC), Cheng additionally outlined case studies where Chinese observational data supported global regulatory and market access activities, including informing FDA discussions and contributing subgroup evidence for European access negotiations. The discussion highlighted how evidence portability is increasingly influencing both development and commercialization strategies across regions.

The Q&A reinforced practical considerations: choosing the right data partner, localizing protocols, understanding registration practices for RWE studies, and criteria for pilot zones like Hainan (unmet need, innovation, and feasibility of local execution).

Beyond Commercially Available Real-World Data: Harnessing Public Data for HEOR Research in the US and Globally

Publicly funded real-world datasets can offer broader population coverage, transparency, and linkage opportunities for HEOR, although speakers in this session highlighted ongoing challenges related to timeliness, access, and clinical granularity. 

Moderated by C Daniel Mullins (University of Maryland School of Medicine), the discussion explored public data sources across the US, UK, and Taiwan. Mullins noted that these datasets are often “more nationally representative” and provide “greater transparency and reproducibility,” while also acknowledging that access processes and data lag can create practical challenges for researchers. 

Lia N Pizzicato (IQVIA) presented a US obesity case study using the Medical Expenditure Panel Survey (MEPS), demonstrating increases in comorbidity burden, healthcare expenditure, and missed workdays across higher body mass index categories. Pizzicato described MEPS as “really well suited for burden on business studies and access analysis,” particularly because of its nationally representative and patient-reported data, while also acknowledging limitations related to self-reported measures and limited clinical detail. 

Focusing on the UK Clinical Practice Research Datalink (CPRD), Cameron C Cook (Amgen) highlighted the value of linking primary care, hospital, and mortality data for cardiovascular research. Cook said linkage is “where the magic happens,” enabling “that near end-to-end inpatient view” across the patient journey. 

Fang-Ju (Irene) Lin (National Taiwan University) discussed Taiwan’s National Health Insurance Research Database (NHIRD) and its linkage with cancer registry and mortality data. Lin explained that the linked databases “create the robust ecosystem” needed to support epidemiology, comparative effectiveness, and policy research at a national scale.

From Engagement to Influence: What Should Patient-Centered Economic Models Actually Look Like?

This Issue Panel session explored how HEOR can move beyond patient consultation toward genuine co-creation of evidence generation and economic assessment. Moderated by Joe Vandigo (Applied Patient Experience, LLC), the session brought together perspectives from Marina Richardson (ICER), Eric W Low (Eric Low Consulting), and Ramiro E Gilardino (R-Impact).

Vandigo highlighted that frameworks already exist to engage patients and capture patient experience data but argued that key challenges remain around how patient-informed outcomes should influence economic assessments, uncertainty analyses, structural assumptions, and equity considerations. Audience polling suggested the greatest current gap is that model outcomes often fail to reflect what matters most to patients.

Richardson presented ongoing patient engagement work at ICER, describing pilot initiatives designed to integrate patient input directly into economic modeling. Examples included revising model structures to better capture disease burden, incorporating caregiver productivity impacts, conducting alternative scenario analyses, and ensuring outcomes better reflected patient priorities and lived experience. Richardson emphasized that while progress is being made, the field is still bridging the gap between where patient engagement is today and where it ultimately needs to be.

Low argued that patient perspectives should not begin at the modelling stage, but much earlier – informing unmet needs, clinical development, and evidence generation from the outset. He stressed that lived experience evidence must be treated with the same seriousness as clinical and economic data and called for stronger investment in generating robust patient experience evidence.

Taking a policy perspective, Gilardino argued that patient engagement must become infrastructure rather than an isolated activity. He highlighted the need for structured, continuous patient involvement across the product lifecycle, suggesting that embedding patient insights earlier could shape endpoints, comparators, and how uncertainty itself is framed.

Across the discussion, speakers agreed that patient-centered economic modelling is not simply about adding patient voices into existing systems – but redesigning systems so those voices help shape evidence from the beginning.