How Much Evidence Proves a Drug Works? Two Trials, One Trial, or One Plus Confirmatory Evidence

A new drug is approved in the United States not because someone believes it works, but because the sponsor has shown "substantial evidence" of effectiveness, a phrase written directly into federal law. What that phrase demands, and specifically how many clinical trials it takes to satisfy it, has been renegotiated over the past three decades and is being rewritten again in 2026.

For most of the modern era the working answer was two. The Food and Drug Administration generally expected two well-designed trials that each pointed to a benefit. A 1997 statute opened a second door: one strong trial plus confirmatory evidence. Draft guidance issued in 2026 now positions that one-trial route as the ordinary path rather than the exception. Understanding why the number changed will teach you something useful about how medical evidence is judged.

Key points#

• The legal standard is "substantial evidence of effectiveness," set out in section 505(d) of the Federal Food, Drug, and Cosmetic Act.
• The FDA long read that standard to require two adequate and well-controlled trials that agree.
• The 1997 FDA Modernization Act (FDAMA 115) allowed one such trial plus confirmatory evidence to count.
• FDA draft guidance in 2023 and a revised draft in 2026 describe what confirmatory evidence looks like and make the single-trial route the default framing.
• These are draft guidances, not law. The statute itself has not changed.

Two questions hidden in one legal phrase#

The core language sits in section 505(d) of the FD&C Act, which defines substantial evidence as evidence drawn from "adequate and well-controlled investigations" that lets qualified experts fairly conclude a drug will do what its labeling claims. That short definition actually asks two different questions.

The first is about quality. An adequate and well-controlled study needs a clear objective, a legitimate comparison group (often placebo or an active drug), design features that limit bias such as randomization and blinding, and an analysis plan set in advance rather than chosen after the data arrive. These features are what let a result be trusted as a real signal rather than an artifact.

The second question is about quantity, and it hides inside a single plural word. Congress wrote "investigations," not "an investigation." From the early 1960s onward, the FDA read that plural to mean two or more. So the familiar two-trial expectation was, in effect, an interpretation of one grammatical choice in a statute.

Why replication became the tradition#

The preference for two trials was not caution for its own sake. It reflects a basic principle that appears throughout science: a single positive result can be a fluke. Chance alone will occasionally produce a favorable trial for a drug that has no true effect. So can an unusual patient sample, a quirk at one study site, or a bias that no one noticed until later.

A useful analogy is measuring something important with a single reading versus taking a second measurement to confirm it. If two carefully run studies, conducted by different teams, land on the same answer, the odds that both were fooled in the same way drop sharply. Requiring a second trial was the regulatory version of insisting the result replicate before betting patients' health on it.

The 1997 turn: one trial plus confirmatory evidence#

That strict reading eased by statute. In the Food and Drug Administration Modernization Act of 1997 (Public Law 105-115), Congress amended section 505(d) to state that the agency may treat "data from one adequate and well-controlled clinical investigation and confirmatory evidence" as substantial evidence when it judges those data sufficient. This provision is usually called by its section number, FDAMA 115.

The goal was not to weaken the effectiveness bar but to acknowledge a reality: sometimes a single, unusually convincing trial, backed by independent supporting data, carries as much conviction as two freestanding studies. That is especially relevant for rare diseases, where enrolling enough patients for even one large trial is hard and a second may be impractical. The statute, however, never defined "confirmatory evidence," which left an open question for years about what would actually satisfy it.

What confirmatory evidence looks like#

The FDA began answering that question in detail in 2023, with draft guidance on demonstrating substantial evidence based on one adequate and well-controlled trial plus confirmatory evidence. The draft lists categories of evidence that can support a single pivotal trial rather than replace it. Here is what you will find on that list:

• Data from a separate, related adequate and well-controlled trial.
• Evidence supporting a closely related indication for the same drug.
• Consistent results from other approved drugs in the same pharmacologic class.
• A well-understood mechanism showing how the drug acts on the disease.
• Natural history or registry data describing how the illness typically progresses without treatment.

The unifying idea is that confirmation can come from somewhere other than a duplicate trial. When a drug's mechanism is clear, the untreated course of the disease is well characterized, and related agents behave predictably, one rigorous trial can sit inside a web of mutually reinforcing evidence. The agency has been explicit that this supporting evidence must be scientifically relevant and reliable, not merely suggestive.

The 2026 shift toward one trial as the default#

The framing moved again in 2026. In June, the FDA issued a revised draft guidance titled "Demonstrating Substantial Evidence of Effectiveness for Human Drug and Biological Products," published in the Federal Register on June 24, 2026, with a comment period into September. When finalized, it is meant to replace the longstanding 1998 guidance. The revised draft centers a single adequate and well-controlled investigation plus confirmatory evidence as the primary route, and again points to sources such as related trial data, class effects, natural history and registry data, and external knowledge of disease pathophysiology.

One caveat matters more than any detail. These are draft guidances, not statute and not final policy. Guidance describes the agency's current thinking; the underlying standard in section 505(d) has not changed, and a draft can shift considerably before it is finalized, so check the status before you cite one.

The trade-off at the center of the debate is genuine and old. Requiring fewer trials can bring effective treatments to patients faster and lower development costs, which matters most for uncommon conditions. The same flexibility loads more weight onto the quality of a single trial and the strength of its supporting evidence, which is why you should read the pivotal trial itself rather than the approval notice. How that balance gets struck, and how consistently across different drugs, is exactly what the current comment period is meant to work out.