ISPOR 2025 daily round-ups: Day 3

As ISPOR 2025 – the leading conference for ISPOR—The Professional Society for Health Economics and Outcomes Research – concludes in Montreal, Canada, we share our final round-up, highlighting key sessions and takeaways from Day 3.

Final Reflections and Leadership Reports 

The closing leadership session of ISPOR 2025 brought together key voices to reflect on the Society’s past, present and future. Rob Abbott (CEO, ISPOR) opened with a vote of thanks, acknowledging the City of Montreal, the conference co-chairs, program committee, ISPOR staff and Board of Directors. He gave special recognition to outgoing board member Brian O’Rourke, praising his leadership in shaping ISPOR’s 2030 strategy. Abbott emphasized the critical need for policy engagement, transparency and trust-building to combat global disengagement in healthcare.

“This is our moment to truly bring health economics and outcomes research to a wider audience.” Rob Abbott

Laura Pizzi (CSO, ISPOR) affirmed that “science is still the priority for ISPOR”, while outlining the Society’s evolving scientific agenda. Pizzi also highlighted ISPOR’s focus on real-world evidence (RWE), artificial intelligence (AI) and health policy, and announced the hiring of ISPOR’s first policy director. Patient engagement was a central theme, with updates on the Patient Initiative Strategy Project and a new patient evidence track launching at ISPOR Europe in Glasgow this year.

Pizzi then welcomed Maarten IJzerman (Erasmus School of Health Policy & Management), Chair of ISPOR’s Health Science Policy Council (HSPC), who described the council’s role in providing independent scientific oversight. He announced that Deborah Marshall would succeed him as Chair and urged members to remain scientifically rigorous and engaged, particularly amid growing global challenges to science and equity.

The remarks reinforced ISPOR’s commitment to advancing science, supporting inclusive dialogue and shaping evidence-based healthcare policy worldwide.

Plenary 3: Balancing Speed and Scientific Rigor―Patient-Centered Methodologies for Surrogate Endpoints in Accelerated Access 

The final plenary at ISPOR 2025 tackled a timely and complex issue: how to balance speed with scientific rigor when using surrogate endpoints to accelerate access to high-need therapies. Bringing together voices from industry, payers and patient advocacy, the session explored how surrogate endpoints (defined as biomarkers or other intermediate outcomes that predict treatment effect on a final clinical outcome) can serve as powerful tools – if they are meaningfully tied to outcomes that matter most to patients.

The panel emphasized the need for early, continuous collaboration in designing evidence strategies, along with deeper patient engagement throughout the lifecycle of medical products. From the evolution of clinical outcome assessments to the ethical imperative of defining “what good looks like” for patients, this plenary underscored that timely access and robust evidence are not mutually exclusive – but achieving both requires thoughtful, inclusive approaches. As the conference concluded, attendees were left with a clear call to action: keep patients at the center, and don’t stop at approval. 

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

Spotlight Session: Cost Benefit Analysis in Value Assessment: Spotlight on Methods 

This session explored the potential role of cost–benefit analysis (CBA) in health technology assessment (HTA), introduced by Eberechukwu Onukwugha (University of Maryland), who outlined the ISPOR HSPC’s mission to address methodological issues through spotlight sessions. She welcomed moderator and discussant Nancy Devlin (University of Melbourne), who set the stage by referencing ISPOR’s strategic focus on Whole Health, which she defined as a framework that acknowledges healthcare’s broader impacts beyond clinical outcomes, including productivity, caregiver burden and social wellbeing. 

Referencing a recent publication in Value in Health, Devlin argued that achieving Whole Health poses challenges for conventional cost–effectiveness analysis (CEA), which tends to focus narrowly on health sector outcomes. She highlighted the potential of CBA to better capture societal value by converting outcomes into monetary terms, facilitating cross-sector comparisons. She noted that while CBA is well established in other public sectors, its use in health remains limited due to both practical and ethical concerns. 

Robert J Brent (Fordham University) delivered the keynote, comparing CEA, cost–utility analysis (CUA), and CBA using dementia-related interventions. He stressed that while CEA helps rank interventions, it cannot determine whether they are socially worthwhile. Brent emphasized that CBA, by placing monetary values on benefits, directly informs whether an intervention merits investment. He illustrated that reliance on budget constraints or cost – effectiveness thresholds can lead to suboptimal or arbitrary decisions. Brent argued that CBA provides a more flexible framework, capable of incorporating diverse outcomes across sectors and accommodating varying budget scenarios. 

Thomas Poder (University of Montreal) responded by acknowledging CBA’s value for allocative efficiency but raised concerns about the reliability of data, especially when derived from stated preference methods. He noted significant variation in willingness-to-pay estimates depending on whether a supply-side or demand-side perspective is taken and emphasized the importance of sensitivity analyses and transparent trade-offs. Poder also highlighted the absence of standardized registries to support CBA in healthcare, unlike in environmental economics. 

Devlin offered a critical perspective on Brent’s examples and reiterated that healthcare markets suffer from market failures such as asymmetry of information and third-party payers. These failures complicate the use of observed prices and necessitate shadow pricing. She concluded that while both CUA and CBA have merits, neither fully addresses equity concerns and both face challenges in operationalization. 

Onukwugha closed the session with a key takeaway: 

“The intuition to move away from a single value, move away from an average, and embrace variability, embrace synthesis methods... provides a really important challenge for us in terms of how we come about quantifying the information that we need to make this operational.” 

This summed up a central message of the session: while CBA offers theoretical advantages, practical implementation requires embracing complexity, heterogeneity and methodological innovation.

Concurrent Sessions 

Wrapping up Day 3 of ISPOR 2025, a series of breakout sessions offered additional learning opportunities.

Workshop: Rare but Common: Generative AI’s Potential on Data, Evidence, and Insight Generation in Rare Diseases 

Generative AI offers new opportunities to address the long-standing challenges of rare disease research, including fragmented data, limited sample sizes and delayed diagnosis. In this workshop, Wayne Su (Jazz Pharmaceuticals) opened by underscoring the scale and urgency of the problem: “Altogether, they're actually not that rare… 300 million people worldwide are living with rare diseases”; yet, “95% have no approved treatments”. 

Xiaoyan Wang (Tulane University) showcased how large language models and knowledge graphs can extract real-world data (RWD) from sources such as electronic health records (EHRs), case reports and social media. She emphasized the potential to generate meaningful insights, stating: 

“We can turn real-time voices into the real, actionable insight.”  

“By using data like the clinical notes, by using technologies such as natural language processing, [we’re] able to get deeper to understand the disease and particular patients,” she explained. 

Jing-Wang Silvanto (Astellas) emphasized AI’s broader role in transforming evidence generation, describing it as “a paradigm shift” and adding, “AI is the only hope, or is the one of the best hopes, for rare disease patients”. She outlined the value of synthetic controls, digital twins and federated learning to support faster, more ethical research and early trial recruitment, stating, “With AI you can create artificial patients… that can make choice more ethical, faster and also much more possible to have success and high power in our submissions”. 

Workshop: Addressing Information Bias in Electronic Health Records and Claims Data: What Can the Literature Tell Us and How Should We Respond? 

This methods-focused workshop, moderated by Patrick Arena (Aetion, Inc.), explored the persistent challenge of information bias in studies using EHRs and claims data for RWE generation. Arena opened by highlighting the growing role of RWE in regulatory and HTA decision-making and emphasized that while EHR and claims data offer rich insights, they were not designed for research – leaving studies vulnerable to measurement error, misclassification and bias.

Allan Meng (Merck & Co. Inc.) presented findings from a targeted literature review conducted with Aetion, which identified 38 relevant publications and synthesized 15 actionable strategies to mitigate information bias. These included data linkage, variable validation and quantitative bias analysis (QBA). Meng illustrated real-world examples where linking EHRs to claims data uncovered exposure misclassification and improved outcome accuracy, and emphasized that a structured mitigation framework is being implemented within Merck’s internal RWE studies.

Daina B Esposito (Moderna) examined information bias through the lens of post-authorization vaccine safety studies, where data completeness and timeliness are critical. She discussed the importance of careful study design, the limitations of claims data in capturing vaccination events, and the role of self-controlled designs and validation in improving study validity.

Finally, Mina Tadrous (University of Toronto) shared insights from the Canadian RWD and RWE landscape, illustrating how linking diverse provincial datasets can uncover bias and improve data quality. Through examples ranging from adherence measurement to off-label drug use and lab data linkage, Tadrous demonstrated that thoughtful cleaning and validation are essential for robust analysis.

The panel emphasized the importance of carefully considering information bias, and that a combination of transparent methodology, collaborative validation and data linkage can significantly improve study reliability.

Workshop: Ensuring the Validity of Real-World Evidence Studies: How Much Can You Check the Data Before You Start?

Moderated by Melvin ‘Skip’ Olson (Olson Strategies GmbH), this workshop tackled one of the issues in RWE research: the ‘gray zone’ of how much data exploration is permissible before finalizing an analysis plan. Olson highlighted the importance of transparency in study protocols and the risks of unintentionally biasing results through overly extensive pre-study data checks.

Helene Karcher (Philip Morris Products S.A.) focused on the dual purpose of data checks – assessing feasibility and orienting analysis design, particularly in studies aiming for causal inference. Using a COPD study comparing heated tobacco product users to smokers, she demonstrated how separate teams performed exposure and outcome checks across distinct datasets to preserve scientific integrity. She also highlighted operational hurdles, including partner reluctance to share data before formal study approval and ethics committee constraints.

Jennifer Christian (Target RWE) introduced a structured solution: a staged approach and clean room committee (described in a recent publication). Drawing from an external control arm study in HER2-amplified biliary tract cancer, Christian explained how a pre-defined, stepwise process can manage sample size limitations, model convergence issues and protocol deviations. The clean room model ensures decisions are documented transparently without revealing outcomes that might bias judgment.

Mary Beth Ritchey (Med Tech Epi; Rutgers University) provided a regulatory perspective, emphasizing the need to assess and communicate uncertainty and bias throughout the study lifecycle. She detailed methods for data validation and operational definitions using real-world examples – such as breastfeeding identification and periprosthetic joint infection diagnosis – to illustrate both feasibility checks and stakeholder expectations.

Together, the speakers emphasized that credible RWE demands not just good data but well-documented, principled processes for handling it – beginning well before the study formally starts.

Issue Panel: Fit-for-Purpose Real-World Data for Medical Device Decision Making: Hype or Hope? 

RWD is increasingly used to inform regulatory and reimbursement decisions for medical devices, yet clear standards for determining whether data are fit for purpose remain lacking. Moderator Eric Barrette (Medtronic) opened by questioning current guidance and underscored the complexity of device evaluation, adding, “Devices and drugs are not the same... a fit for purpose assessment might get much more complicated, very quickly.” 

Farah Husein (Canada’s Drug Agency) emphasized the importance of consistent, transparent standards for evaluating RWE. “In all cases, there needs to be transparency around what was done, how it was done, and contain enough contextual information to understand and interpret the results,” she said. Husein argued that:  

“The real-world evidence standards for devices do not need to be different from those that are required for drugs.” 

Ami Buikema (Optum) highlighted the fragmentation of device data across the healthcare system. “The device codes do not actually usually make it into electronic health records,” she noted, adding, “we are putting our pieces together from a variety of different data sources”. She pointed to the missed potential of unique device identifiers, stating, “That’s really unfortunate, because there’s an actual identifier that you can use to find the device, yet it’s not being captured consistently or across the healthcare system in a way that makes it easily usable for research purposes. And more importantly, it can be used for things that are really, really important to patient safety and quality, like product recalls or safety post-authorization”. 

Katherine Mues (Aetion) stressed the importance of rigorous and transparent methods in RWE generation. “You need clear and well documented data provenance, application of principled epidemiologic methods and complete transparency on the methods and results and how those were developed,” she explained. Mues noted that progress will depend on “more standardization in how studies are designed, how data is assessed, and how ultimately, a regulatory or payer submission looks.”

Issue Panel: Are Social Determinants of Health (SDoH) Data Ready for Primetime?  

In a debate-style session, Smita Kothari (Merck Sharp & Dohme International Service BV) opened by defining social determinants of health (SDoH) as the environmental conditions in which people live, work and age, which are typically grouped into domains such as education, economic stability and healthcare access. She noted the sharp rise in SDoH-focused literature and implementation efforts by health systems, setting the stage for a discussion on whether SDoH data are truly “ready for primetime”. 

Starting with the opposing view, C. Daniel Mullins (University of Maryland School of Pharmacy) argued that SDoH data remain largely hypothetical in their utility. Drawing on a study conducted with Ontada in a community oncology setting, Mullins highlighted gaps in evidence linking SDoH factors to key clinical endpoints like progression-free survival. He cautioned against over-reliance on proxies such as ZIP code-based indices, warning of potential implicit bias and the flawed assumption that removing a social determinant will automatically eliminate poor outcomes. “We assume that because there's a strong correlation between a social determinant and a negative outcome, removing that determinant will make the outcome go away – that’s not necessarily true,” Mullins stated. 

Amy K O'Sullivan (Ontada) countered with evidence from several Ontada studies. In one study, socioeconomic deprivation and race were associated with lower rates of genetic testing among patients with triple-negative breast cancer. Another large-scale analysis linked race and deprivation to later-stage diagnoses and shorter time to metastasis across eight cancer types. O’Sullivan also highlighted the implementation of the NCCN Distress Thermometer in EHRs, which helped identify unmet social needs and connect patients to support services. “The data may not be perfect, but they’re good enough to generate meaningful insights,” she concluded, urging greater transparency and provider-patient trust in data collection efforts.