The two-trial era may be over. The hard part isn't.

Adigens Health helps sponsors design evidence strategies that meet the standard; from target trial specification to decision-ready confirmatory evidence packages. Get in touch at info@adigenshealth.com.

The Food and Drug Administration (FDA) does not often rewrite its own operating assumptions. When it does, the consequences ripple for years. A sounding board published in the New England Journal of Medicine in February 2026, authored by FDA leaders, did exactly that: it announced that one adequate and well-controlled clinical trial, combined with confirmatory evidence, will now be the FDA's default standard for marketing authorisation. The two-trial rule is, in their words, a dogma whose time has passed.

The argument has a coherent logic. Modern drug development is not what it was in 1997, when the statutory flexibility for single-trial approval was first formalised. Mechanistic science has become more precise. Biomarkers have multiplied. The FDA now considers biochemical and intermediate changes alongside survival data, looking for what the authors call "a complete biologic story." In that richer evidential landscape, demanding a second pivotal trial regardless of context is, they argue, a blunt instrument. A second poorly designed trial adds little. A single rigorously designed one, surrounded by mechanistic credibility, can be decisive.

There are also the economic facts. A single pivotal study costs between $30 million and $150 million. Drug development already exceeds seven years on average. If that burden can be halved (and if drug prices are partly defended on R&D cost grounds) then reducing it removes at least one element of an argument that has long irritated payers and patients alike.

The authors are not naïve about the risks. They acknowledge, directly, that the FDA has approved drugs under the two-trial standard that later proved unsafe or ineffective. The number of studies, they note, is no safeguard if trial design is deficient: a substandard control arm, a post-hoc statistical plan, inadequate power all of these can undermine two studies as surely as one. The proposal, they insist, does not lower the bar. It redirects attention toward what actually determines evidential quality.

The FDA has declared the two-trial era over. The industry should welcome the change. It should also be careful what it wishes for.

The permission is not the plan

That case is well made. But it leaves open a question the paper does not answer: if one trial is to carry the weight of approval, how exactly should it be designed to bear it? This is not a rhetorical question. The history of single-arm and externally controlled studies is littered with credible-looking answers that unravelled on closer inspection. The confirmatory evidence that is meant to support a single trial can mean many things (animal models, class effects, surrogate biomarkers, real-world data) and some of those things are substantially more credible than others. Without a disciplined framework for how the pivotal trial relates to all the evidence around it, the one-trial default risks becoming a one-shortcut default.

The discipline that resolves this problem has a name. The target trial framework, originally articulated to improve observational research, provides a structured way to specify, before data generation begins, what a study is actually trying to estimate. Population. Treatment strategies. Assignment procedure. Follow-up. Outcomes. Estimand. Causal contrast. These are the architecture of a valid inference, and every weakness in that architecture becomes a weakness in the approval decision.

Confirmatory evidence needs confirmatory rigour

The paper introduces a second concept that deserves equal scrutiny: confirmatory evidence. Under the new default, a single trial does not stand alone; it sits within a broader evidentiary picture. That picture can include mechanistic science, data from related indications, class-level information, or real-world evidence. When coherent, this evidence package is genuinely reassuring. The example the authors cite (imatinib in chronic myeloid leukaemia, whose population-level mortality benefit became visible in SEER data) is a genuine case where the confirmatory evidence was so strong that the single-trial approval looks, in retrospect, almost conservative.

But imatinib is not the modal case. Most drugs approved will produce modest effect sizes in surrogate endpoints, surrounded by real-world data of variable provenance. In that more typical situation, the quality of the confirmatory evidence matters enormously. Quality, in real-world data, is not assumed. It is designed.

The target trial framework provides the required discipline. Emulating a target trial with real-world data means specifying, explicitly, the hypothetical randomised trial the observational analysis is meant to approximate: the same population, the same treatment strategies, the same follow-up rules, the same outcomes. It forces clarity on time zero. This deceptively simple concept, when handled poorly, produces immortal-time bias and results that do not replicate. It demands a bias register rather than a limitations paragraph: confounding, misclassification, informative censoring and practice variation named, assessed, and mitigated before the analysis is run.

Sponsors who treat real-world confirmatory evidence as a narrative asset, something to round off a regulatory submission rather than a piece of causal reasoning, will find that the FDA's new scrutiny on study quality catches up with them. The authors are explicit: the agency will focus its finite reviewer time on the one trial it requires, examining controls, endpoints, effect size, and statistical protocols with renewed attention. Confirmatory evidence that cannot withstand the same scrutiny will not compensate for a pivotal trial that cannot stand alone.

What this means in practice

The new default changes the planning calculus for sponsors in several concrete ways. The target trial protocol should be written early, ensuring the confirmatory evidence strategy is coherent with, not merely adjacent to, the pivotal trial design. The comparator must be contemporary. The FDA's approval record includes cases where novel drugs were compared against treatments no longer used by American patients. Under a one-trial standard, with reviewer attention concentrated rather than spread, that kind of comparator will be harder to defend. The target trial framework forces this question early: what would a well-designed trial actually use as the control arm?

The choice of confirmatory evidence should be governed by causal logic, not availability. Mechanistic data that elucidates a precise, non-redundant pathway is genuinely confirmatory. Real-world data analysed without target trial discipline is not. Sponsors should identify, in advance, which evidence sources meet the standard and which do not. Finally, the FDA reserves the right to require two studies when a mechanism is nonspecific, an outcome labile, or a trial design deficient. That reservation is significant. Sponsors who approach the new default as an invitation to do less will discover it is, in the FDA's framing, an opportunity to do better. The agency expects a surge in drug development. What it has not promised is a surge in approvals.

The bottom line

The FDA's shift from two trials to one is a sensible response to how drug development has changed. It does not, however, solve the underlying problem of how to extract a valid causal inference from limited evidence. That problem remains, and it requires a framework.

Target trial thinking is that framework. Sponsors who apply it rigorously, to the pivotal study and to the confirmatory evidence around it, will be well positioned for the new standard. Those who treat the one-trial default as a cost-saving shortcut will find they have traded one set of requirements for a harder set of questions.

One trial is enough. The question is always whether it was the right one, designed the right way, surrounded by evidence that earns its place. The FDA has raised the stakes. The discipline to meet them already exists.