The Accelerated-Approval Bargain: Surrogate Endpoints, Confirmatory Trials, and Faster Withdrawal

Accelerated approval is a promise made in two installments: a drug for a serious illness reaches patients now, on the strength of an early laboratory signal, and the maker owes proof of real benefit later. The U.S. Food and Drug Administration can clear such a drug based on a surrogate endpoint that is "reasonably likely to predict clinical benefit," rather than waiting for direct evidence that people live longer or feel better. The catch is written into the deal: the sponsor must run a confirmatory trial, and if the benefit never shows up, the approval can be pulled. The 2022 Food and Drug Omnibus Reform Act, widely called the 2023 accelerated-approval reform, sharpened both halves of that contract. It let FDA insist the confirmatory trial already be running at approval, spell out its terms, and unwind an unverified approval through a quicker, defined process.

Key points#

• Accelerated approval grants market access early, based on a surrogate rather than a proven clinical outcome.
• The price of that early access is a mandatory confirmatory trial to verify the benefit.
• A surrogate can improve while the outcome patients care about does not, which is the built-in risk.
• FDORA (2022) let FDA require the confirmatory trial be underway at approval and created faster withdrawal procedures.
• The core standard, "reasonably likely to predict clinical benefit," did not change; the enforcement around it did.

Surrogate versus clinical: what the drug is actually measured on#

Start with the two kinds of yardstick. A clinical endpoint captures something a patient directly lives through: survival, symptom relief, fewer hospital stays, less disability. A surrogate endpoint is a stand-in, usually a lab value, an imaging finding, or a physical sign that is expected to move in step with the clinical outcome. A tumor that shrinks is a proxy for a longer life. A falling biomarker is a proxy for a slowed disease.

The appeal of a surrogate is speed. It can be read out sooner and in a smaller group of patients, so a treatment for a life-threatening condition can reach you years ahead of a trial powered on survival. The statute makes room for exactly this: FDA may approve on a surrogate "reasonably likely to predict clinical benefit," or on an intermediate clinical endpoint that shows up earlier than irreversible harm or death (FDA, Accelerated Approval Program). The phrase "reasonably likely" is chosen with care. It sits well below "established," and it encodes a judgment made under uncertainty, on the view that for a grave illness, waiting also carries a cost.

That gap between "likely to predict" and "proven to deliver" is not a defect in the design. It is the design. A biomarker can shift in the hoped-for direction while the patient gains nothing, which is why the pathway is structured as approval with strings attached rather than approval and done.

The obligation that comes with the access#

Granting accelerated approval triggers a duty on the sponsor: run a post-approval confirmatory trial built to verify and describe the expected effect on a genuine clinical outcome (FDA guidance, Expedited Programs for Serious Conditions). Meanwhile the drug is on pharmacy shelves and reaching patients, yet the very question it was approved on stays formally unanswered until that trial reports.

From there, three futures are possible. The trial confirms benefit, and the drug converts to a standard, unconditional approval. The trial fails to confirm benefit, and the approval becomes eligible for withdrawal. Or, the outcome that caused most of the trouble historically, the trial simply grinds on for years while the drug remains available on the surrogate alone. When a confirmatory study started late or stalled midway, the "conditional" in conditional approval could outlast its intended shelf life without anyone forcing the issue.

Where FDORA tightened the screws#

Enacted in December 2022 inside the Consolidated Appropriations Act, FDORA amended section 506(c) of the Federal Food, Drug, and Cosmetic Act. Rather than touch the approval standard, it reinforced the machinery around it at each weak joint.

The trial must already be moving#

FDA gained explicit authority to require, where appropriate, that a confirmatory study be underway before approval is granted, or within a set window afterward. This attacks the delay problem at the root. A promise to open a trial someday is replaced by an expectation that patients are already enrolling as the drug launches. A draft guidance carried in the Federal Register in January 2025 laid out the signals FDA weighs when judging whether a trial is genuinely "underway," among them a firm target completion date, demonstrated progress, and active patient enrollment (FDA draft guidance, Determining Whether a Confirmatory Trial Is Underway). Being draft, it reflects FDA's proposed reasoning rather than a settled rule.

The terms are fixed at the start#

FDORA directs FDA to specify the conditions of the required studies up front, which can cover enrollment targets, interim milestones, and completion dates. Locking these down at approval leaves less room to argue later about what the sponsor owes and by when.

Progress is reported more often, in public#

The law raised the tempo of accountability. Sponsors must report on required post-approval studies at least every 180 days, and FDA publishes that status. A stalled trial becomes visible sooner when its progress is on the record.

Withdrawal has a faster lane#

The most consequential shift is at the exit. FDORA built expedited withdrawal procedures for an accelerated approval, usable when a confirmatory trial fails to verify benefit, when a sponsor does not pursue a required trial with due diligence, or when other statutory grounds apply. The streamlined route keeps procedural fairness intact: written notice with reasons, a chance for the sponsor to meet with the agency, possible advisory-committee input, and publication of the proposal for comment. The point is a removal mechanism that is quicker and better defined than the older, cumbersome one, without denying the sponsor a hearing.

A council to keep the bar even#

FDORA also asked FDA to stand up an internal Accelerated Approval Program Council to keep the pathway applied consistently across review divisions, with public reporting on what it does. It is an institutional guard against drift, so one division does not hold a looser or stricter line than another.

Why this trade-off resists a clean answer#

The tension here is not a bug awaiting a fix; it is the reason the pathway exists at all. Demand more certainty before approval and patients with serious disease wait longer for treatments that might help them, and some will not have the time. Approve on thinner surrogate evidence and some drugs reach the market whose promised benefit never lands, leaving you with cost and side effects for an uncertain payoff. Accelerated approval packs that dilemma into one structure: earlier access, bought with a binding duty to prove the benefit, backed by a real way to reverse course.

What FDORA changed is how firmly the second installment can be collected. The surrogate standard, "reasonably likely to predict clinical benefit," is untouched. What is different is how tightly the confirmatory duty is set at the front end and how efficiently an unverified approval can be undone at the back end. Call any single approval too hasty or too cautious as you like; the underlying shape is the same bargain, speed now, in return for proof later and a credible lever to act when the proof does not come.