The FDA's postmarket study system rests on a promise: a drug or device can reach patients before every safety question is answered, as long as the manufacturer answers the remaining questions afterward. The new KFF Health News analysis of the FDA's own database shows how often that promise goes unkept. Nearly 600 postmarket studies were behind schedule as of August, covering almost 350 products. More than 250 of those studies had final reports originally due before July 31. About a third of all ongoing studies were flagged as delayed, and separate FDA device databases list dozens more.
The numbers describe a system where the deadline is the least important part of the requirement. A study can miss its date by years, the product stays on the market, and the only consequence is a row in a public database that moves from one overdue category to the next. The patients who take these drugs and use these devices are the ones who were promised answers, and the answers are not late by accident. The system is built to tolerate lateness, and the incentives point that way.
The database is the story
KFF's analysis is a database story, and its power is that the database is the FDA's own. The agency publishes postmarket requirements in a tracker that anyone can read, and the analysis simply asked what the tracker says about deadlines: which studies are overdue, by how much, and for which products. The answer required no leaks and no insider accounts. The public record shows hundreds of studies past due, some by more than a decade.
That transparency is worth pausing on, because it complicates the easy narrative. The FDA does not hide these delays; it publishes them, product by product, study by study. The problem is not secrecy but follow-through. A regulator that discloses its backlog but has weak tools to clear it is a regulator whose disclosure becomes the documentation of a commitment it cannot enforce, and every year of published overdue studies makes the gap between the promise and the practice more visible.
Tavneos is what the tail looks like
The example the analysis leads with is the one that shows where this road ends. Tavneos, an Amgen drug for a rare inflammatory disease, was approved in 2021 with a required safety study of 300 patients. As of the fall, 21 were enrolled. The FDA later identified dozens of cases of drug-induced liver injury, including deaths, and has moved to withdraw the drug's approval, alleging that pivotal trial results were manipulated. The European Union revoked the drug's authorization, and the New England Journal of Medicine retracted the supporting study. Amgen disputes the allegations, and the drug remains on the market while the dispute runs.
The case is an outlier in its severity and a pure specimen of the system's logic. The postmarket study was the condition that made the original approval acceptable. The study barely enrolled. The drug kept selling. Only after safety signals accumulated outside the study did the withdrawal machinery start, years later. The postmarket promise is supposed to catch the Tavneos cases early. Instead, Tavneos became the case that shows what the promise cannot do when the required study stalls.
The lesson generalizes past one drug. Every overdue study is a Tavneos-shaped risk at lower amplitude: a product whose residual uncertainty is supposed to be shrinking but is not. The database cannot say which of the hundreds of overdue studies hides a genuine safety problem, and that is exactly the point. The system's promise was that patients would not have to guess which uncertainties were being resolved. When the tracker shows the resolution date sliding year after year, the guessing shifts back to the patients, their doctors, and the market.
Delay pays, and the system knows it
Harvard's Aaron Kesselheim put the incentive in one phrase quoted by KFF: delay can pay, because the product stays on the market while the study is pending. The arithmetic is simple. Finishing a confirmatory study costs money and risks a negative result. Not finishing it costs, in most cases, nothing measurable. A company facing those two columns will resolve most conflicts in the same direction.
The money trail supports the point. The HHS Office of Inspector General estimated that Medicare and Medicaid spent more than $18 billion from 2018 through 2021 on accelerated-approval drugs whose confirmatory trials were incomplete past their deadlines. Sarepta's Exondys 51, a muscular dystrophy drug, missed its confirmatory study deadline in May 2021 and remained delayed while generating $538 million in sales last year. A required study that is overdue is, for the balance sheet, a problem solved. That is the central finding of the whole analysis, and it is not an accusation against any single company. It is a description of the board the FDA built and the way the game on it is played.
The reverse case completes the picture. Oxaydo, an abuse-deterrent opioid approved in 2011, had a required study due in 2016. The FDA issued a failure-to-respond letter in 2022, six years later, and the product has since been discontinued. The lesson is not that the FDA eventually acts. It is the gap: a decade of marketing between the missed deadline and the end of the product, with the required evidence never delivered. When the punishment for a missed deadline arrives after the product's commercial life is effectively over, the deadline was not a deadline.
The honest counterpoint deserves its weight
The other side of the ledger is real, and a fair account has to carry it. An HHS spokesperson told KFF that delays can have legitimate causes and "should not be treated as evidence that a product has an unresolved safety or effectiveness issue." Some studies stall because enrollment is genuinely hard, because the science moved, or because a better question replaced the original one. A JAMA Network Open study of 370 drugs approved from 2016 through 2023 found that 86.4 percent of required trials were completed by the end of 2023 and 88.2 percent reported results, with safety trials taking a median of 5.4 years against 3.3 for efficacy trials.
Those findings do not excuse the backlog; they bound it. Most required studies do eventually finish, and most of the system works. The problem is that the exceptions are concentrated exactly where the stakes are highest: accelerated approvals, safety studies, and products whose confirmatory evidence is weakest at the moment of approval. A system that completes most of its studies can still be a system in which the ones that stall are the ones that mattered most, and the Tavneos file shows the tail is not merely long. It is where the damage lives.
The device backlog is the quieter one
The drug numbers dominate the coverage, but the analysis found the same pattern on the device side, where it gets even less attention. The CustomFlex Artificial Iris had a pediatric study protocol accepted in 2019, and as of August the study had enrolled zero patients. The Scandinavian Total Ankle Replacement, approved in 2009, had enrolled 142 of a required minimum of 500 subjects, and nearly half of those enrolled had experienced adverse events. Paxlovid's pregnancy safety study missed its end-of-2024 deadline, a gap that matters because the drug's pregnancy use is one of the questions prescribers most often ask about.
Device postmarket studies carry the same logic as drug studies with a thinner public record. Device approvals move faster, the required studies are smaller and less standardized, and the products are implanted in patients rather than dispensed, which makes a stalled safety study harder to notice and harder to undo. The KFF analysis's device findings are the part of the backlog least likely to produce headlines and most likely to persist, because nobody is watching the tracker for an artificial iris the way they watch it for a billion-dollar drug.
The single-trial default stacks the backlog
The analysis lands at a moment when the load on the system is set to grow. In February, the FDA announced that its default approval requirement would drop from two clinical trials to one, a change framed as accelerating access. The arithmetic for postmarket requirements is unavoidable: if drugs reach the market on less pre-approval evidence, the post-approval studies carry more of the evidentiary weight, and a system already carrying hundreds of overdue studies would be asked to carry more. UCSF's Sanket Dhruva told KFF that postmarket requirements have been "often toothless," and the single-trial change makes the teeth matter more, not less.
The fix is not a mystery, and it is not more requirements. It is a clock with consequences. Regulators have the tools in principle: accelerated approval exists precisely so that withdrawal can follow a failed or missing confirmatory study. What the database shows is that the tools are used rarely, slowly, and late. Until an overdue study has a price attached to it, the rational company will treat the deadline the way the tracker says hundreds of companies already have: as a suggestion.
Primary sources
- KFF Health News (David Hilzenrath), whose September 18 analysis of the FDA postmarket databases supplies the figures, the Tavneos account, and the expert commentary.
- MedPage Today, for its summary of the KFF findings and the context of the FDA's requirements system.
- Becker's Hospital Review, for the HHS response and the JAMA Network Open completion statistics.