There is a moment in the movie “The Fugitive” when Dr. Richard Kimble, cornered at the lip of a dam with nowhere left to run, tells Deputy US Marshal Sam Gerard the only thing that matters to him. “I didn’t kill my wife,” he says. Gerard, weapon raised, delivers the line that made the movie: “I don’t care.”
Amgen may know the feeling. The company insists the pivotal efficacy data behind its vasculitis drug Tavneos were not manipulated, arguing that a 2019 reanalysis by the original sponsor followed a prespecified procedure, altered no underlying patient data, and does not amount to the deception the US Food and Drug Administration alleges.
The agency’s position, stated with US marshal gruffness, is that intent is almost beside the point, because once the reliability of a single pivotal trial is thrown into doubt, it is very hard to reassert. And whatever did or did not happen in that reanalysis, it happened at ChemoCentryx in 2019, three years before Amgen bought the company.
For the uninitiated, “The Fugitive” follows Harrison Ford as Dr. Richard Kimble, a surgeon wrongly convicted of murdering his wife, who escapes custody and spends the movie both dodging a relentless US Marshal and hunting the one-armed man who he says actually did it. The 1993 movie has a 96% Tomatometer score from critics and features perhaps the most dramatic scene ever set at a medical conference. Check it out!
Like Kimble, Amgen is protesting its innocence over conduct it did not itself commit, to a pursuer more interested in the record than the plea.
A Manhunt On Four Continents
The pursuit of the drug is now global. The most definitive action came on Aug. 4, when the European Commission revoked Tavneos’ marketing authorization after the European Medicines Agency concluded the study data used to support the application were “incorrect and misleading” and could no longer be relied upon. CSL Vifor, which markets the drug in Europe, said it was disappointed but would implement the decision in full.
Across the Atlantic, the FDA acted first, but the process is moving more slowly. In an April 27 letter, the agency proposed withdrawal on two grounds: that new information, when weighed against what was known at the time of approval, shows a lack of substantial evidence of effectiveness, and that the NDA contained untrue statements of material fact. The agency’s specific charge is that unblinded personnel at the original sponsor readjudicated endpoint results for nine ADVOCATE patients, flipping a not-significant result into a significant one, and that the readjudication was never disclosed. The New England Journal of Medicine retracted the ADVOCATE paper in June for the same reason.
Other regulators are still loading their weapons. The UK’s Medicines and Healthcare products Regulatory Agency (MHRA) opened a benefit-risk review in June after the data integrity questions surfaced, and Japan’s Ministry of Health, Labour and Welfare issued a reporting order against licensee Kissei over its handling of the trial results. In Australia, the Therapeutic Goods Administration said this month it was reviewing information that has raised concerns regarding the integrity and reliability of the clinical trial data that supported its approval of Tavneos.
For a small biotech watching this unfold, the lesson is not subtle: a single pivotal trial with an integrity question attached is a single point of failure, and one regulator’s doubt now travels across borders at the speed of a shared PDF.
Watchdog has never been accused of under-doing analogies, so we should note that there is an almost too-perfect irony buried in “The Fugitive”: the crime Kimble ultimately exposes is a pharmaceutical company burying a drug's liver toxicity to protect its approval. In the real-world withdrawal fight over Tavneos, liver safety hovers over everything, though regulators are not suggesting there was a cover-up about it.
The One-Armed Man Is More Common Than You Think
If the phrase “manipulated data in an acquired application” sounds like a freak occurrence that you hear about at the movies instead of the boardroom, it is not. The trouble is that the inherited fraud plot usually surfaces long after the closing dinner, when the buyer is holding the asset and, therefore, the liability.
The canonical case is Daiichi Sankyo’s roughly $4.6bn purchase of Ranbaxy in 2008. A whistleblower and an internal self-assessment revealed fabricated data used to win approvals in more than 40 countries. The fallout ran for years: import bans, a 2012 consent decree, and a 2013 guilty plea to seven felony counts paired with a $500m settlement. Daiichi eventually sold Ranbaxy to Sun Pharmaceutical at a loss, then won a Singapore arbitration award after a tribunal found the Ranbaxy selling shareholders had concealed the investigations.
The distinction from Tavneos is that Ranbaxy’s problems were with manufacturing and bioequivalence integrity across its portfolio, not with a single efficacy trial.
The closer structural cousin is Novartis and AveXis. Novartis paid $8.7bn for AveXis in 2018, its gene therapy Zolgensma was approved in May 2019, and the FDA later found the application contained manipulated data from animal studies used to establish the manufacturing process (with the added wrinkle that AveXis knew before approval and told the agency only afterward).
Here, the ending diverged sharply from where Tavneos may be heading. The FDA characterized the tainted data as a small, preclinical slice that did not undermine the clinical evidence, kept the drug on the market, and closed its investigation without penalties, while Novartis parted ways with the researchers involved. That “small portion, clinical data intact” framing is precisely the escape hatch Amgen does not have, because the alleged manipulation in ADVOCATE strikes the sole pivotal efficacy trial, which is why the FDA can argue there was never substantial evidence at all.
Even Kimble-Level Diligence Can Miss It
The most uncomfortable case for anyone who believes a thorough audit can avoid problems is Akorn v. Fresenius. In 2017, Fresenius agreed to buy the generics maker Akorn for $4.75bn, then discovered data integrity problems and walked. In October 2018, the Delaware Court of Chancery held (for the first time ever) that a material adverse effect entitled the buyer to terminate. It is the case every deal lawyer now cites, and it should worry every diligence team, because Fresenius did the work and still missed the fraud.
Fresenius conducted detailed due diligence, and its own management deck described Akorn’s level of access as above average for a public-company target. Yet the pre-signing review flagged competition risk and an under-resourced regulatory affairs group, while concluding that no deal breakers had been identified. The concealment was deliberate: during the gap between signing and closing, Akorn’s most senior quality official submitted a Complete Response Letter reply he knew would put fabricated data before the FDA, specifically to avoid a red flag that might jeopardize the sale before it closed. You cannot easily audit your way past a counterparty who is actively hiding evidence.
What actually saved Fresenius was not diligence but architecture. Akorn had made extensive FDA compliance representations qualified by a material adverse effect standard, plus a covenant to operate in the “ordinary course” between signing and closing, and a covenant granting Fresenius reasonable access to its records.
When whistleblower letters arrived after signing, Fresenius used that access covenant to run a real investigation, staffed with a former acting FDA chief counsel and a technical firm to test the underlying data. Akorn, by contrast, conducted a probe with a litigation partner who had never undertaken a data integrity investigation or appeared before the FDA. The contrast was not lost on the court, and it should not be lost on you.
Three Unglamorous Fixes
An important lesson from these cases is that pre-signing diligence is necessary and insufficient, and the government agrees. The Department of Justice’s Evaluation of Corporate Compliance Programs (most recently updated in 2024) treats acquisition scrutiny as diagnostic of your whole compliance function. Prosecutors are told to ask:
- whether you completed pre-acquisition due diligence and, if not, why not;
- whether misconduct or its risk was identified;
- who ran the review;
- how findings were tracked and remediated; and
- whether you audit newly acquired entities after closing.
Done well, that scrutiny lets you shift the cost of a target’s misconduct onto the seller. Done badly, it lets the misconduct (and the liability) ride along into your building.
So the practical steps are three, and none of them is the kind of thing you would see in an action thriller. First, negotiate the backstop before you fall in love with the asset: specific FDA-compliance representations qualified by a material adverse effect standard, an ordinary-course covenant with teeth, and broad post-signing access rights are what let a buyer walk when the diligence turns out to have been fooled.
Second, staff the integrity question with people who have actually done integrity work (former agency counsel and technical data auditors), not the deal team, because a litigation partner who has never read a CGMP guidance cannot tell a prespecified analysis from a cooked one.
Third, take your own consultants’ warnings seriously. Akorn had been rated among the three worst data-integrity performers out of 120-plus companies that one assessor had reviewed, but then the company curtailed the very audits that would have confirmed it.
An Exciting Third Act Twist!
Which brings us, at last, to Amgen’s defense, and to the plot twist it is banking on. The company argues that the 2019 readjudication followed a prespecified procedure applied to unaltered patient data, that no underlying clinical values were changed, and that the FDA has not identified an actual untrue statement of material fact. It’s asking the agency to reverse course, or at least put the withdrawal question to an advisory committee.
To back up that request, Amgen commissioned a fresh, blinded readjudication by the Duke Clinical Research Institute, which found Tavneos noninferior to a prednisone taper at week 26 (68.1% versus 67.1%) and on sustained remission at week 52 (61.4% versus 52.4%), with a nearly 10-point numerical edge at week 52, lower relapse, and more than 80% less steroid exposure.
Superiority was not demonstrated, which Amgen concedes, but noninferiority with a fraction of the glucocorticoid toxicity is not nothing for a rare disease with roughly 3,500 US patients relying on the drug.
The stone the builders rejected, though, is the real-world evidence. Amgen’s package also leans on seven comparative effectiveness studies, more than 20 observational cohorts, an ongoing Japanese postmarketing surveillance study, and a European observational study, all offered to show the trial result was not a mirage. And here is the delicious twist for anyone who has sat through a decade of conference agendas.
Real-world evidence (RWE) was the hot acronym not so long ago, the belle of every panel, the thing that was going to modernize evidence generation. Then artificial intelligence walked in, and RWE got quietly moved to the kids’ table while everyone crowded around the new arrival. Regulators, for their part, have long treated real-world data as supporting evidence rather than the star of the show.
So picture the Hollywood ending. The character everyone underestimated, the sidekick who was only ever supposed to corroborate, turns out to be the one who clears the wrongly accused defendant and catches the one-armed man. If a body of real-world evidence talks an advisory committee into keeping Tavneos on the US market after its pivotal trial was retracted and its EU authorization revoked, RWE will have done something its promoters spent years promising and rarely delivering: saved the day when it counted.
Do not cue the swelling strings yet, however, because the FDA has already boxed in one of Amgen’s two arguments. On the readjudication question, the agency has telegraphed a good counter: “Every additional readjudication or reanalysis reduces the reliability of the result because of the increased risk of chance findings.” In other words, the more times you reanalyze to prove the drug works, the more the agency suspects you are fishing.
The FDA has not reckoned with the real-world evidence, because it arrived with Amgen’s request for a hearing after the agency’s withdrawal proposal was written. Kimble had to produce the one-armed man in the flesh. Amgen has to convince the FDA, and possibly an advisory committee, that its supporting cast is a leading man. Whether that is a triumphant third act or a straight-to-streaming disappointment, we will find out at the hearing – assuming the FDA grants one.
See you tomorrow.