In 2011 the US Food and Drug Administration approved Makena, an injection meant to reduce the risk of preterm birth in women who had previously delivered early. It went through the accelerated approval pathway, which lets the agency approve drugs for serious conditions on the basis of a surrogate or intermediate measure, on condition that the manufacturer later confirms a real clinical benefit.
The confirmatory trial, known as PROLONG, was completed in 2019. It found no benefit in preventing preterm birth. The FDA proposed withdrawing the drug in 2020. The withdrawal was finalised in 2023.
Twelve years on the market. Almost four of them after a confirmatory trial had shown it did not work.
The debate about regulatory speed is usually framed as a trade-off: approve new therapies faster and patients get access sooner, but safety suffers; approve slowly and patients wait, but get more certainty. That framing makes the approval decision the centre of everything. My argument is that it misses where the real risk sits. Fast approval, done openly, is a defensible bet. The problem is that regulatory systems are built to say yes quickly and to say “we were wrong” very slowly. The asymmetry is the danger, not the speed.
What accelerated approval is supposed to do
The FDA created its accelerated approval programme in 1992, largely in response to the HIV crisis, when patients and activists argued, with good reason, that waiting years for mortality data meant waiting to die. The logic was sound. If a drug clearly changes a biological marker that is reasonably likely to predict benefit, such as viral load or tumour shrinkage, it can be approved early, while trials measuring the outcomes that matter to patients are completed.
Much of the time, it works. Drugs approved this way have included treatments that later proved their worth in full trials. For patients with serious diseases and no alternatives, earlier access to something that turns out to work is an enormous benefit, and one that is easy to overlook when attention focuses on the failures.
But the bargain has two halves. Early approval is the first. Rapid, reliable confirmation, or rapid withdrawal if confirmation fails, is the second. The first half has been delivered consistently. The second has been much weaker.
The asymmetry, in three cases
Makena shows how long withdrawal can take even after a failed trial. Two other cases show different sides of the same problem.
In June 2021 the FDA approved aducanumab, sold as Aduhelm, for Alzheimer’s disease under accelerated approval, on the basis of its effect on amyloid plaques in the brain. The decision was controversial. The agency’s own advisory committee had not recommended approval, and three advisers resigned after the decision. The European Medicines Agency rejected the drug in December 2021. A report by two US House committees in 2022 criticised the FDA’s interactions with the manufacturer. In January 2024, Biogen abandoned the drug.
Aducanumab was not withdrawn because a confirmatory trial failed. It faded because doctors, payers and patients did not trust the evidence and did not use it. The market reached a verdict faster than the regulatory process did.
Voxelotor, a treatment for sickle cell disease, received accelerated approval in 2019. It was among the first new sickle cell drugs approved in the US since 1998, which made it significant for a patient community that has long been underserved. In September 2024 Pfizer pulled it from markets worldwide, after further data raised concerns about its safety. As an analysis in Blood Advances noted, crizanlizumab, another sickle cell treatment, had already been removed from the European market in August 2023. For patients who had been told these were breakthroughs, the reversals damaged trust in the whole system.
In each case, the approval moved quickly. The correction, where it came, was slow, messy or driven by someone other than the regulator.
The mechanism: why confirmation stalls
Why does the second half of the bargain fail so often? The answer lies in incentives that change the moment a drug is approved.
Before approval, a manufacturer has every reason to run trials quickly: no approval, no revenue. After accelerated approval, the drug is already on the market and earning money. A confirmatory trial can only do one of two things: confirm what the company is already selling, or threaten it. The incentive to finish quickly weakens.
Recruitment gets harder too. Once a drug is approved and available, patients and doctors are less willing to enter a trial where they might be randomised to placebo or an older treatment. Why join a trial when you can get the drug from a pharmacy? Confirmatory trials can struggle for years to enrol enough patients.
Regulators, meanwhile, face their own asymmetry. Approving a drug for desperate patients is a popular decision. Withdrawing one, even an ineffective one, means confronting patients who believe it helps them, doctors who prescribe it, advocacy groups who campaigned for it and a manufacturer with lawyers. Before the law changed, withdrawal in the US required a lengthy formal process. Every incentive pointed toward delay.
The reform that has already happened
The US Congress recognised part of this problem. The Food and Drug Omnibus Reform Act, passed at the end of 2022, gave the FDA power to require confirmatory trials to be under way before accelerated approval is granted, and established expedited procedures for withdrawing approval when those trials fail or are not conducted with due diligence.
That is the right direction. It addresses the recruitment problem by starting trials before approval, and the withdrawal problem by shortening the process. What remains to be seen is how consistently the agency uses those powers, especially when withdrawal is politically uncomfortable. Legal authority is necessary but not sufficient. The agency also needs the staff to monitor hundreds of post-approval commitments, and the political backing to act when a well-known drug with a vocal patient community misses its deadlines. Powers that are rarely used tend, over time, to stop being feared.
A scenario: the regulator downstream
Consider a mid-sized regulator in a middle-income country with a small assessment team. It cannot run a full independent review of every new medicine, so, like many agencies, it relies on decisions by reference regulators to speed up its own approvals. That reliance is sensible. Duplicating a large agency’s work would waste scarce expertise and delay access for patients.
Now suppose a reference regulator grants accelerated approval to a new cancer drug on the basis of tumour shrinkage in a single-arm trial. The smaller regulator follows within months. The drug is added to hospital formularies. Patient groups lobby for it to be covered by the national insurance scheme, and it is, at a price negotiated with little leverage.
Three years later, the confirmatory trial fails to show that patients live longer. The reference regulator begins the withdrawal process. But the smaller regulator’s approval was not conditional on the confirmatory trial, because its own rules had no such mechanism. Nobody in the agency is assigned to track confirmatory trials for drugs approved through reliance. The drug stays on the local market, and in the insurance benefit package, long after it has gone elsewhere.
This is not hypothetical in its structure, even if the details are illustrative. Reliance pathways import approvals quickly. They rarely import withdrawals with the same speed. For health systems in lower-income countries, which can least afford to pay for drugs that do not work, the asymmetry is multiplied.
The fix is administrative rather than scientific: approvals granted through reliance on an accelerated decision should carry the same conditions, and someone should be responsible for watching what happens to the evidence.
The strongest counterargument
The strongest argument for leaving drugs on the market while evidence accumulates is that withdrawal can harm patients too. If a drug is helping some people, even if a trial shows no average benefit, taking it away may hurt them. Trials measure average effects in selected populations. Individual patients and their doctors may see benefits that trials miss.
There is also a concern about chilling innovation. If companies believe approvals will be reversed quickly at the first sign of trouble, they may be less willing to invest in drugs for rare or difficult diseases, where evidence is hardest to generate.
These concerns have some force. Individual variation in response is real, and patients with rare diseases have legitimate reasons to fear regulatory caution. But they argue for well-designed confirmatory trials and clear rules, not for leaving ineffective drugs on the market indefinitely. A system that cannot remove drugs that have failed their confirmatory trials is not protecting patients. It is protecting the approval decision.
The innovation concern also cuts the other way. When high-profile accelerated approvals fail, public and professional trust in the pathway falls. That makes it harder for the next genuinely effective drug approved on early evidence to be taken up. Credible withdrawal protects the value of fast approval.
Second-order effects
The asymmetry between approval and withdrawal has consequences beyond individual drugs.
It shifts costs to payers. Public health systems and insurers often have to pay for drugs approved on limited evidence, at full price, for years before confirmation. Aducanumab’s launch price was reported at US$56,000 a year per patient, and Medicare had to decide how to cover it. Coverage with evidence development, where payers fund drugs only while data are collected, is one response, but it adds another layer of decision-making.
It affects regulators elsewhere. Many countries rely heavily on FDA and EMA decisions when deciding what to approve. A fast US approval can lead to approvals elsewhere, while a slow US withdrawal may not be matched quickly by other regulators. The aducanumab case, where Europe said no, shows that regulators can diverge, but many smaller regulators lack the capacity to make independent assessments.
It shapes patient trust. Patients who are told that a drug is a breakthrough, and later learn it was withdrawn, may become more sceptical of future approvals, including good ones. Trust, once lost, is expensive to rebuild.
What this means for each actor
- For regulators: treat withdrawal as a core function, with clear timelines and public reporting on the status of every confirmatory trial. Make the second half of the bargain as visible as the first.
- For health ministries and payers: link coverage and pricing of drugs approved on early evidence to the completion of confirmatory trials. Pay less until benefit is confirmed.
- For regulators in smaller countries: avoid automatically following a single reference regulator. Track withdrawals and divergent decisions, not just approvals.
- For clinicians: be candid with patients about the evidence behind early approvals. “Approved” does not always mean “proven.”
- For patient advocates: the fight for fast access and the fight for fast confirmation are the same fight. Weak confirmation eventually undermines access.
What to watch next
Watch how often the FDA uses its expedited withdrawal powers in the coming years, and how long withdrawals take in practice. Watch whether confirmatory trials are genuinely under way at the time of accelerated approval, as the law now allows the agency to require. And watch whether other regulators and payers adopt similar rules, so that the speed of approval is matched, everywhere, by the speed of correction.
Speed was never the enemy of safety. A regulator that can move quickly in both directions is safer than one that moves slowly in both. The one to fear is the regulator that is fast only when saying yes.
Should drugs approved on early evidence be priced lower until their benefit is confirmed? Tell us what you think.
