A new medicine arrives at a regulator's desk with two kinds of proof attached. The first is the familiar kind: a randomised controlled trial, blinded, controlled, expensive, slow, and narrow in whom it enrolled. The second is newer and looks, on paper, like everything the first is not — drawn from electronic health records, insurance claims and disease registries, assembled in months rather than years, and reflecting the messy reality of patients as they actually present in clinics across Asia. The promise of that second kind of proof is enormous. The question every regulator in the region is now wrestling with is deceptively simple: when is it reliable enough to decide on?

This is the unresolved tension at the centre of real-world evidence (RWE) in the Asia-Pacific. It is not a debate about whether RWE has a place — that argument is effectively over. It is a debate about where the line sits between evidence that contextualises and evidence that decides, and APAC's regulators are drawing that line in real time, market by market, submission by submission, with frameworks that are still mostly provisional and acceptance standards that are still mostly case-by-case.

The pull: evidence that is faster, cheaper, and finally inclusive

Start with why RWE is irresistible, because the pull is real and it is not merely commercial.

The first force is speed and cost. A pivotal randomised trial is the most rigorous instrument medicine has, and also one of the slowest and most capital-intensive. For a region where regulators are under pressure to close the lag between a drug's approval abroad and its availability at home, evidence that can be generated from data already being captured in routine care is structurally attractive. Claims databases, electronic medical records and registries hold years of patient histories that cost nothing additional to create. The marginal expense is analysis, not enrolment.

The second force — and the one that matters most for the Asia-Pacific specifically — is representation. Classical trials are built on inclusion and exclusion criteria that produce clean cohorts and, in the process, exclude the patients clinicians actually treat: the elderly, the multi-morbid, the patients on five other medications, and, persistently, Asian populations themselves. A drug developed and trialled predominantly in North American and European patients arrives in Seoul, Mumbai or Jakarta carrying an implicit question about whether its effect holds in a different genetic, dietary and care-pathway context. Real-world data generated inside the region is the most direct answer to that question. It is local evidence about local patients, and for regulators charged with protecting their own populations, that local quality is not a nice-to-have. It is the point.

The third force is feasibility in places where a randomised trial is impractical or unethical — rare diseases with too few patients to randomise, oncology indications where withholding a promising therapy is hard to justify, and conditions where the natural history is so well-characterised that a contemporaneous comparison from real-world data can stand in for a control arm. Here RWE is not a cheaper substitute for a trial. It is sometimes the only evidence available at all.

Put those three forces together and the appeal is obvious: faster decisions, lower cost, and — uniquely valuable in APAC — evidence about the actual populations a regulator is responsible for. The trouble is that every one of those advantages comes attached to a liability.

The reliability problem: bias, data quality, and a missing rulebook

Real-world data is generated for purposes that have nothing to do with answering a regulatory question. A hospital records a diagnosis to bill for it. A clinician notes a lab value to manage a patient, not to power a statistical comparison. When that data is later repurposed as evidence, the gap between why it was collected and what it is now being asked to prove becomes the source of every serious objection.

The headline problem is confounding. In a randomised trial, the act of randomisation distributes both the known and the unknown differences between treatment groups, so that a difference in outcome can be attributed to the drug. Real-world cohorts are not randomised. The patients who received a new therapy may differ systematically from those who did not — sicker, healthier, younger, better-insured, treated at a more specialised centre — in ways the data captures imperfectly or not at all. Statistical methods can adjust for the differences that are measured. They are powerless against the ones that are not. "Unmeasured confounding" is the phrase that recurs, almost ritually, in every regulatory rejection of an RWE package, and it is the structural weakness no amount of sample size can fix.

Beneath confounding sit the more mundane but equally corrosive problems of data quality: missing values, inconsistent coding between hospitals and between countries, diagnoses recorded with varying precision, outcomes that are ascertained differently depending on who was looking. Two key tests recur in the guidance that has emerged — whether the data is relevant, meaning rich enough to actually answer the clinical question being asked, and whether it is reliable, meaning accurately and consistently captured at the point of collection. A dataset can be enormous and still fail both.

And then there is the deepest problem, which is not technical at all: the absence of a settled framework for when an answer is good enough. Regulators broadly agree on the principles — high-quality sources, eligibility criteria matched to the clinical question, contemporaneous data, representative populations, transparent statistics to handle confounding. Where they have not converged is on the threshold. How much residual bias is tolerable? When does an external control arm built from registry data carry the same weight as a randomised one? At what point does observational evidence stop being supportive context and start being substantial evidence of effectiveness? There is no equation for that line, and so it is being negotiated submission by submission, which is precisely what makes RWE strategy so fraught for sponsors and so consequential for regulators.

The cautionary cases are instructive, and APAC regulators watch them closely even when they originate elsewhere. In November 2025, the US Food and Drug Administration rejected Biohaven's troriluzole for spinocerebellar ataxia despite a regulatory package that paired a placebo-controlled study with a three-year real-world study that had met its efficacy endpoints; the agency's complete response letter cited the bias, design flaws, lack of pre-specification and unmeasured confounding it described as potentially inherent to real-world and external-control studies. Y-mAbs' omburtamab for neuroblastoma drew a negative recommendation over the comparability of its external control arm and its choice of time-to-event outcomes. Even where one agency accepts an RWE-supported case, another may not: avapritinib in advanced systemic mastocytosis was supported in part by real-world data serving as an external control, and while it cleared its regulatory hurdles, Germany's reimbursement assessor declined to consider the same evidence, judging the documentation and the indirect-comparison methodology inadequate. The lesson is consistent. The same dataset can be decisive in one jurisdiction and inadmissible in the next, because the line is not yet drawn in a common place.

Regional moves: how APAC is building its own answer

What makes the Asia-Pacific distinctive is that its regulators are not waiting for a global consensus to form. Several are building the infrastructure, the guidance and the case law themselves — and they are doing it in markedly different ways.

China has been the region's most assertive mover, and its central instrument is unusual: a place. Since 2018, the Boao Lecheng International Medical Tourism Pilot Zone in Hainan has operated as the only zone in China where patients can be treated with drugs and devices approved overseas but not yet registered domestically — and, crucially, where the real-world data generated from that use can be submitted to support national registration. It is a regulatory sandbox engineered specifically to manufacture RWE for products that meet urgent, unmet needs. By mid-2022, two dozen licensed imported drugs and devices had been brought into the pilot. The National Medical Products Administration (NMPA) has built guidance around the model, beginning with its 2020 draft guideline on using real-world data to support the clinical evaluation of medical devices and extending into drug-specific procedures for the Lecheng pathway. Beyond the sandbox, China has demonstrated cross-border RWE in action: an expanded osteoporosis indication for denosumab was approved on the mainland in June 2020 supported in part by real-world evidence drawn from Taiwan's National Health Insurance Research Database and Hong Kong's Clinical Data Analysis and Reporting System, using data on Chinese patient populations to substantiate effectiveness in routine practice. It is one of the clearest examples in the region of observational evidence carrying genuine regulatory weight.

Japan has taken the opposite temperament — methodical, registry-centric, and built on reliability above all. Rather than leaning on broad electronic-health-record or claims data, the Pharmaceuticals and Medical Devices Agency (PMDA) has invested in highly curated sources: the MID-NET distributed database, originally built for pharmacovigilance, and a deliberate emphasis on disease registries as the preferred vehicle for regulatory-grade real-world data. PMDA established a dedicated registry consultation in 2019 and issued foundational guidance in March 2021 on the basic approach to using registries in approval applications and on ensuring their reliability. The trajectory is visible in the numbers. Of 674 new drug and regenerative-product applications approved between 2019 and 2024, 158 — roughly 23 per cent — incorporated real-world data or evidence, and the share climbed steadily across the period, from around 18 per cent of applications in 2019 to over 30 per cent in 2024. That is no longer experimental; it is becoming routine, and it brings Japan to a level of RWE utilisation comparable to Europe's.

Across the rest of the region the picture is a mosaic. South Korea's MFDS and its allied health-data infrastructure have been active particularly in medical devices, where short product life cycles and thin trial evidence make real-world data a practical necessity. Singapore's Health Sciences Authority operates in one of the region's richest and most centralised health-data environments, giving it unusual capacity to generate and interrogate real-world data, while the city-state has also become the intellectual hub of the region's RWE thinking. Australia's TGA has engaged with real-world data largely through post-market surveillance and registry-based monitoring. India's CDSCO sits earlier on the curve, with the principles acknowledged but a formal acceptance framework still maturing.

The most important regional development, though, sits slightly to the side of drug approval itself. The REALISE working group — REAL World Data In ASia for HEalth Technology Assessment in Reimbursement — was established in 2019 under HTAsiaLink, led from the National University of Singapore's Saw Swee Hock School of Public Health and Thailand's Health Intervention and Technology Assessment Program, and brings together experts from eleven Asian health systems. Its output is a non-binding guidance document aimed at aligning how the region generates and uses real-world data — not for marketing authorisation, but for the reimbursement and health-technology-assessment decisions that determine whether an approved drug is actually paid for. This matters in APAC because the regulatory and reimbursement layers are tightly coupled and sometimes simultaneous; evidence judged adequate to approve a drug may be judged inadequate to fund it, and vice versa. REALISE represents the region's attempt to build a shared vocabulary and a shared set of expectations before the divergence between eleven separate national approaches hardens into eleven separate rulebooks.

Underneath every one of these regulatory moves sits a quieter but decisive question of infrastructure: a regulator can only accept real-world evidence as robust as the platform that produced it. This is where the region's variation is starkest. Taiwan's National Health Insurance Research Database and Hong Kong's Clinical Data Analysis and Reporting System are mature, near-complete population datasets precisely because they were built on single-payer systems that capture almost every encounter — which is why they were credible enough to help carry the denosumab indication across a national regulator's threshold. Japan's MID-NET is smaller and more curated, deliberately engineered for reliability rather than scale. Singapore's centralised health records give its agencies an unusually clean evidentiary base for a high-income, data-dense system. In much of the rest of the region, by contrast, the data is fragmented across hospitals, payers and private providers, coded inconsistently and stitched together with difficulty. The uncomfortable implication is that the markets best positioned to use real-world evidence are the ones whose health systems already capture it well — and that the populations who would benefit most from local evidence are sometimes those whose data infrastructure can least support it. Closing that gap is as much a regulatory priority as writing the guidance, because guidance without a trustworthy platform underneath it decides nothing.

That impulse toward harmonisation has a global counterpart. The International Council for Harmonisation endorsed a reflection paper in 2023 on aligning real-world evidence terminology and the general principles for planning and reporting studies that use real-world data. It is early and it is conceptual, but it signals that the question APAC regulators are answering nationally is one the entire regulatory world has agreed needs a common answer.

Designing for credibility: the discipline that earns the decision

If there is a single conclusion emerging from the cases where RWE has succeeded and the cases where it has failed, it is this: the credibility of real-world evidence is decided long before the data is analysed. It is designed in or designed out at the planning stage.

The first principle is fit-for-purpose. Real-world evidence is not a single thing of uniform strength; its acceptability depends entirely on what it is being asked to do. As a source of safety signals and post-market surveillance, it is now broadly accepted across the region — Japan's MID-NET-based safety monitoring is a mature example. As context for a single-arm trial in a rare disease, it can be persuasive when the natural history is well-understood. As the primary basis for a claim of comparative effectiveness against an active competitor, it faces the steepest scepticism, because that is the claim randomisation was invented to support. Matching the ambition of the evidence to the strength of the design is the first discipline, and overreaching is the most common way RWE packages fail.

The second principle is methodological transparency and pre-specification. The recurring complaint in rejections is not only bias but the suspicion of bias introduced after the fact — analyses chosen once the results were known, endpoints adjusted, cohorts redefined. The antidote is to lock the protocol before looking at outcomes: pre-specify the study objective, the eligibility criteria matched to the comparable trial population, the outcomes, the statistical approach to confounding, and the sensitivity analyses, and then report all of it transparently. An external control arm built from registry data can attenuate certain biases through careful design — contemporaneous data, harmonised eligibility, appropriate adjustment — but only if those choices are made and declared in advance rather than reverse-engineered toward a desired result.

The third principle, and the one sponsors consistently underweight, is early and continuous engagement with the regulator. Because there is no settled acceptance threshold, the threshold is effectively negotiated. The packages that succeed are the ones where the sponsor aligned with the agency on data source, comparator, endpoints and statistical framework before the study began — not the ones that arrived at the desk hoping the evidence would speak for itself. In a region where each regulator is still forming its own standard, that conversation is not a courtesy. It is the mechanism by which the standard gets made.

Where 'real' carries weight — and where it doesn't yet

So where does this leave a sponsor, a regulator, or a reader trying to gauge how much regulatory weight real-world evidence can actually bear in the Asia-Pacific?

The honest answer is that it depends on the question, the market and the design — and that the answer is moving. For safety and post-market surveillance, RWE is firmly established across the region's leading agencies. For rare diseases, oncology and other settings where randomisation is impractical, it can support primary regulatory decisions, particularly through well-constructed external control arms, and China's Lecheng pathway and Japan's registry framework show two different but functioning routes to exactly that. For local validation — confirming that an effect demonstrated elsewhere holds in Asian populations — real-world data is arguably the region's most valuable evidentiary asset, and its strategic importance will only grow.

Where RWE still cannot reliably carry the decision is at the hardest edge: standing alone as substantial evidence of effectiveness for a novel therapy where a randomised trial was feasible and simply was not run, or supporting a comparative-effectiveness claim against an active competitor on the strength of an observational cohort. There, the unmeasured-confounding objection remains decisive, and APAC regulators — watching the same rejections their Western counterparts are issuing — are not yet willing to draw the line further out.

The frameworks are provisional, the thresholds are unsettled, and the region is harmonising in fragments. But the direction is unambiguous. Real-world evidence in the Asia-Pacific has moved from the margins of the regulatory file to a permanent place within it. The remaining work is not to decide whether 'real' counts. It is to decide, case by case and then principle by principle, exactly how much — and on that question, APAC is no longer waiting for the rest of the world to answer first.

BioPharma APAC — Next-Generation Clinical Operations Series. This feature is part of an ongoing editorial examination of how the region's manufacturing and clinical-development functions are being reshaped under pressure.

DISCLAIMER

This feature is an independent editorial analysis prepared by the BioPharma APAC editorial desk. It synthesises publicly available regulatory guidance, agency decisions and peer-reviewed literature current as of June 2026. Regulatory frameworks for real-world evidence are evolving rapidly; acceptance standards differ by market and may change after publication.

Named approvals, rejections, data platforms and guidance documents are drawn from public records and are cited illustratively to characterise the landscape, not to represent the position of any agency, sponsor or institution. Figures are rounded and reflect the sources available at the time of writing.

Nothing in this article constitutes legal, regulatory, clinical or investment advice. Readers making submission or strategy decisions should consult the relevant national regulator and qualified advisers.

© 2026 BioPharma APAC.