Real Growth: In Health Technology Assessment, Real-World Evidence Is Thriving

Health technology assessment bodies are increasingly relying on real-world evidence to inform decisions that impact payment and access for medical treatments. How will health economics and outcomes research help this approach realize its potential?

By Beth Fand Incollingo

More and more, when evaluating urgently needed treatments that don’t lend themselves to standard clinical trials, health technology assessment (HTA) bodies and payers are turning to real-world evidence (RWE) for answers. But while RWE’s ability to inform such decisions is past its infancy, experts say it has more to offer.

“I do not believe that we have yet realized the full potential of real-world evidence,” said Grace Li-Ying Huang, Senior Director of the Division of Health Technology Assessment at Taiwan’s Center for Drug Evaluation. “In many ways, we are still at the beginning of this transformation. Over the past decade, much of the focus has been on building databases and developing analytical methods. Over the next 5 years, however, I expect real-world evidence to become a core component of healthcare infrastructure.”

Over the past decade and a half, HTA bodies have increasingly incorporated RWE into their decisions about value, market access, and reimbursement as a supplement to data from clinical trials, and now decision makers from the United States and the European Union to Asia-Pacific and Latin America are encouraging it through a network of policies and guidelines.

Consider the regulatory approvals of Kymriah (tisagenlecleucel), a one-and-done chimeric antigen receptor T-cell (CAR T) therapy that can spark long-lasting remissions in patients with certain blood cancers but has a 6-figure price tag. The US Food and Drug Administration, the European Commission, and Health Canada gave the therapy regulatory approval based on single-arm clinical trials with relatively short follow-up periods, leaving HTA bodies to decide whether or how to support it. Italy and Spain responded by building registries that generated RWE about Kymriah’s effectiveness—information HTA experts in those countries needed to strike timed-payment and risk-sharing agreements with the drug’s manufacturer.

RWE also helped the National Institute for Health and Care Excellence (NICE) evaluate the cost-effectiveness of pre-exposure prophylaxis (PrEP) for HIV prevention in the United Kingdom. By evaluating UK-specific RWE, NICE determined the anticipated patient need, uptake estimates, and health outcomes, ultimately finding the strategy worthwhile for citizens at high risk of contracting HIV.

Especially in the United States, private health insurers are responding to this evolution by including RWE in their assessments, and pharmaceutical companies across the world are cooperating by building RWE into their drug development life cycles in efforts that can range from creating patient registries to conducting long-term, postlaunch follow-up studies.

The next hurdle, experts say, will involve ensuring that the RWE submitted for these assessments is credible, complete, relevant, and reproducible.

Finding Real-World Value

Perhaps even more than regulators, HTA professionals appreciate real-world findings because they help determine when treatment patterns differ from clinical guidelines or controlled study environments, whether safety signals match what trials demonstrated, and how patients perceive the therapeutic experience, said Dimitra Lambrelli, MPharm, PhD, a London-based Senior Director and Senior Research Scientist on the PPD^TM Evidera^TM, Real-World Data Scientific Solutions Team at Thermo Fisher Scientific, a life sciences and laboratory technology company.

HTA bodies can act on that information by matching costs to value through conditional reimbursement mechanisms and managed entry agreements or by imposing postlaunch restrictions on reimbursement or patient eligibility, Huang added.

Most often, experts say, HTA decision makers use RWE to assess the value of precision drugs, gene therapies, or other innovative treatments for rare diseases—for which randomized clinical trials are not always possible and approval processes may be accelerated. RWE is also frequently used to amass long-term data about safety, survival, and health outcomes in cancers or chronic conditions such as cardiovascular disease or obesity. Finally, HTA bodies may consider RWE when weighing the potential value of a label expansion for a drug.

The next hurdle, experts say, will involve ensuring that the real-world evidence submitted for health technology assessments is credible, complete, relevant, and reproducible.

The mission of ensuring that RWE is reliable for these purposes will fall to experts in health economics and outcomes research (HEOR), whose unique skill set includes gathering and studying data harvested from electronic health records, medical insurance claims, administrative databases, disease and product registries, and patient-generated sources such as surveys, forums, and wearable devices.

HEOR professionals also have expertise in applying advanced statistical strategies known as causal inference methods to help minimize the effect of confounding factors, such as differences between populations being studied, missing information, or inconsistencies in the way real-world data has been collected.

“To bring credibility, HEOR leaders must clearly understand the need to produce high-quality studies that meet payers’ stringent demands,” said Sebastian Schneeweiss, MD, ScD, a Professor of Medicine and Epidemiology at Harvard Medical School and Chief of the Division of Pharmacoepidemiology at Brigham and Women’s Hospital, both in Boston in the United States. “Real-world evidence is not a shortcut to success, and high-validity studies will get us further. Following proven guidances and principles will get us to the right place.”

Shaping Robust RWE

While many of today’s techniques for conducting real-world research were developed by academics, they’re largely used by pharmaceutical companies seeking to inform decisions by regulators and HTA bodies, as well as by national health systems.

One key framework is target trial emulation, which enables researchers to design the randomized clinical trial they wish they could conduct and then use real-world data to create an observational study mimicking that structure. Applying causal inference methods can make it possible for researchers to meaningfully compare estimated outcomes among their emulated comparator cohorts.

A related method involves complementing a single-arm clinical trial with an external control arm. Mirroring a trial’s endpoints and populations, these real-world cohorts gather patient data from electronic health records and medical charts. One notable external control arm helped provide NICE with evidence that Tecentriq (atezolizumab) could spark long-term survival in some patients with non-small cell lung cancer. That study, which supplemented the OAK trial with a real-world control cohort of patients treated with docetaxel, was created by Roche using data from the Flatiron patient database.

Causal inference frameworks are crucial in this strategy, too, to ensure that trial arms can be compared and to help prevent bias or systematically missing data, which Lambrelli said can be a challenge when using RWE. For instance, since patients typically pay for GLP-1 obesity treatments out of pocket, those transactions are missing from some key databases, such as insurance claims systems.

“A current issue in our field is the applicability of the right methodologies,” she said. “While they need to become more accessible and more appropriately used, these approaches require a lot of statistical understanding, and their success depends on how knowledgeable the scientists are who apply them.”

As part of that effort, HEOR experts working with RWE must consider why data were originally collected—for research, administrative oversight, commercial activities, or patient support—and whether they were gathered systematically and transparently, Lambrelli added. While data not generated for research, such as conversations in online patient forums, can be valuable, they need to be handled differently. For instance, “social media listening” studies can highlight the patient perspective as a way of setting the scene for more quantitative research.

But scientists must keep that information as pristine as possible, too.

“It’s best to pick up sources that contain a lot of information and then use artificial intelligence (AI) to run sensitivity analyses that identify outliers,” such as “trolls” who comment with the aim of provoking or disrupting, Lambrelli said. “The appropriate methodologies applied to harvest this information are equally important to the quality of the data itself.”

Unfortunately, the credibility of real-world studies varies widely, Schneeweiss said.

Recently, a number of questionably designed studies have been published by medical students who pulled information from TriNetX, a database of anonymized electronic health records that is freely available to researchers at participating healthcare organizations. Many of these papers are misleading and can leave regulators and HTA decision makers confused, Schneeweiss said.

“With these studies poisoning the water,” he said, “nobody knows what to trust anymore.”

Researchers can combat that problem by prioritizing transparency about their protocols and processes—one tactic involves preregistering study designs for consideration by the scientific community—and by using established methods to quantify measurement characteristics, such as accuracy, reliability, and completeness.

HEOR professionals who rely on proven frameworks to evaluate the quality of their protocols are doing exactly what they should, Schneeweiss said, but now the field needs to seek international agreement around just a handful of these tools to bring more consistency to real-world research.

“Better reliability will come as we provide more education and training programs and harmonize and test these assessment tools with the input of regulators and HTAs,” he said.

Health economics and outcomes research experts working with real-world evidence must consider why data were originally collected and whether they were gathered systematically and transparently.

Leading the Effort

HTA bodies in the United Kingdom, Scotland, Canada, France, and Germany are among those that incorporate RWE into their decisions, using it to better understand patients and their medical needs and to assess the cost-effectiveness of treatments after they’ve hit the market.

Beyond simply evaluating that evidence, though, some countries and regions are leading efforts to shape how RWE is collected and applied.

These initiatives include the public-private GREG project in the European Union, which aims to help stakeholders generate consistently reliable RWE for use in HTA decisions; Canadian Real-World Evidence for Value of Cancer Drugs (CanREValue), which is producing a framework for the use of RWE in funding decisions about oncology drugs; and a proposal by academics from ISPOR’s Taiwan chapter to establish a framework that would help HTA experts define and assess the value of expensive new oncology, orphan, and other drugs whose treatment efficacy remains uncertain.

Focusing on RWE is not unusual in Taiwan, where HEOR experts have developed several efforts that are changing the way the country makes value decisions about medical treatments, with more under consideration, Huang said. Those initiatives include:

• A Cancer Drugs Fund. Informed in part by Taiwan’s exchanges and collaboration with NICE International, this mechanism allows promising cancer drugs that address important unmet medical needs to receive temporary reimbursement despite remaining clinical uncertainty. During the temporary reimbursement period, additional postlaunch evidence, including RWE, may be collected or submitted to assess clinical effectiveness, safety, utilization, and budget impact. This evidence can then inform health technology reassessment and subsequent reimbursement decisions.

• A national registry of patients who take immune checkpoint inhibitors, which supports reimbursement decisions and label expansions for those drugs.

• The international Fast Healthcare Interoperability Resources standard, which Taiwan is applying across 5 oncology areas with the goal of making it easier for the country’s healthcare organizations to exchange real-world data.

• Consent forms that are now routinely requested from patients to allow the collection and analysis of their health information. This effort aims to ease concerns about privacy that can sometimes stand in the way of data exchange.

• Proposed guidance for drug developers about the RWE components they are expected to include in HTA review packages. Taiwan hopes to publish a final version by the end of 2026.

Taiwan is particularly well positioned to conduct real-world research, Huang said, because its National Health Insurance program covers more than 99% of the population and provides access to nationwide healthcare databases that may be linked for research under appropriate data-governance arrangements. Nevertheless, researchers must apply rigorous causal inference methods to address confounding, selection bias, and other methodological challenges. They must also carefully assess and manage data-quality issues, such as potential misdiagnosis or overdiagnosis, coding inconsistencies, incomplete documentation, and fragmented patient records across different healthcare institutions,
Huang said.

Some low- and middle-income countries, such as India and Brazil, have found it much more challenging to incorporate RWE into HTA decisions, largely because they lack access to high-quality real-world data. That can occur due to immature digital infrastructures or disconnected systems for sharing health information.

“We like to say that your real-world evidence is as good as your real-world data, so if you don’t have those data or appropriate methodologies for assessment, there’s not much you can do,” Lambrelli said.

She added that while privacy policies can sometimes restrict the sharing of real-world information, the application of data governance techniques, patient de-identification approaches, and AI strategies “should allow all the players to work together to ensure the privacy of individuals while at the same time granting wider access to good-quality data.”

Maximizing RWE’s Potential With AI

Not surprisingly, AI is transforming the generation of RWE used in HTA decisions by enabling better data extraction, quality, and anonymization, along with predictable analytics and nearly real-time health outcomes surveillance, experts said.

“Through natural language processing and machine learning, we’re able to use data that we could not have analyzed before—notes in patients’ medical charts, pathology reports, imaging narratives, laboratory findings, insurance claims, patient registries, and information reported by wearable devices,” Lambrelli said. “AI can also help us assess the quality of that data and whether it’s fit for purpose by identifying missingness, coding inconsistencies, duplicate records, and outliers.”

That can benefit HTA bodies by clarifying therapy sequencing patterns and the incidence of medicine switching in the real world, as well as by predicting the likelihood of disease progression, hospitalization, treatment response, therapeutic adherence, reasons for drug discontinuation, relapse, and adverse events.

AI can also save money, preserve privacy, and reduce the need for vast real-world collection of information by providing synthetic data, sometimes even creating “digital twins” that can stand in for patients in certain research settings, Lambrelli said. Limiting the involvement of real patients can be especially useful in rare-disease research, she said, where trial populations are typically small and easily identifiable through context clues.

Despite the gains these functions can bring, experts caution that HEOR professionals must remain in the loop to supervise AI’s activities, checking in at various points to make sure they can see and understand the processes it’s using to gather information. Otherwise, they can end up with a black-box model, which may produce statistically accurate data but offers no transparency about how outputs are connected to inputs.

Artificial intelligence is transforming the generation of real-world evidence for health technology assessment by enabling better data extraction, quality, and anonymization, along with predictable analytics and nearly real-time health outcomes surveillance.

“AI is already very helpful in extracting unstructured information and turning it into structured information, which then can feed into the analytic causal inference pipeline,” Schneeweiss said. “There are published agentic AI tools to write and evaluate protocols, turn them into analyses, and interpret data. The key things that consumers of this information need to consider are how much they want to remain in control and how much transparency they need from AI agents and tools to make decisions about big populations and big budgets.”

Moving Toward Maturity

Moving forward, Huang expects HEOR professionals across the drug development process to not only continue generating traditional cost-effectiveness analyses, but to increasingly integrate RWE into evaluations of healthcare utilization, long-term value, patient quality of life, health equity, caregiver burden, and broader societal outcomes.

For RWE to become a core component of healthcare evaluations, Huang said, stakeholders will need to tackle complex projects designed to improve methodology, data interoperability, ethical guidelines, and AI applications—all of which will require international cooperation.

In Taiwan, we are eager to collaborate with partners from different countries,” Huang said. “We value opportunities both to share our experiences and to learn from others. This kind of collaboration can help all of us use real-world evidence more effectively as part of robust and transparent HTA and reimbursement decisions.”

Beth Fand Incollingo is a freelance writer who reports on scientific, medical, and university issues.