NEJM Retracts Avacopan

An unprecedented retraction reveals cracks in the very foundations of drug development

On June 29^th of 2026, the New England Journal of Medicine (NEJM) retracted the ADVOCATE study, which described a phase 3 randomized controlled trial of the C5a receptor inhibitor avacopan (Tavneos) as compared to a prednisone taper. This marks the first time in the >200-year history of NEJM it has retracted a registrational study that served as the basis for regulatory approval. It also marks the first time it retracted a rheumatology trial.

This will be a pivotal moment for rheumatology, but I hope folks outside of rheumatology are also paying attention. This is not just a story about avacopan. It is a story about how the normal systems of sponsor oversight, contract research organization (CRO) data handling, peer review, regulatory approval, and post-publication scrutiny failed to detect a shocking case of data manipulation.

Re-Re-adjudicating Avacopan

For a full rundown of what happened, please read my recent substack on Re-adjudicating Avacopan or grind your way through the 24 page notice of opportunity for a hearing documentation that lays out the case in exhaustive detail. Here’s a brief summary:

ADVOCATE was a 52-week randomized controlled trial that pitted avacopan against a prednisone taper in patients with ANCA associated vasculitis. The company that owned avacopan at the time, ChemoCentryx, reported positive headline results in November of 2019, and the drug was approved in October of 2021. In January of 2026, the FDA asked for it to be withdrawn voluntarily. Amgen, which had acquired ChemoCentryx in 2022, declined.

In April the FDA published the notice of opportunity for a hearing (NOOH) that lays out its case. According to the FDA, when ADVOCATE was designed, the FDA told ChemoCentryx that approval would require superiority at week 52. After enrollment finished, the database was unblinded and two company employees, the chief medical officer (Dr. Pirow Bekker) and the director of biostatistics (Dr. Huibin Yue), saw the results. The critical week 52 superiority outcome had not been met (p = 0.1025).

What happened next is the part the FDA calls data manipulation. Rather than report the null result, the agency alleges, Bekker and Yue worked backward from the answer they wanted. They ran a post-hoc analysis that showed a mere 5 patients flipping their outcome would change the overall results of the trial (the “reverse fragility index” for those interested in this topic). They then identified 9 patients for blinded re-adjudication, which for complicated reasons could only land in avacopans favor. It worked; after re-adjudication, the headline result became statistically significant.

When ChemoCentryx filed for approval, they left out the original (clean) analysis, did not disclose that the data had been re-analyzed, and did not disclose the re-adjudication. The FDA’s notice disclosing these facts is unusually blunt. It accuses Bekker and ChemoCentryx of data manipulation, material omissions, and untrue statements of material fact.

This is serious, but it was also just one notice from one agency. When I wrote my last article, the FDA’s European counterpart (the EMA) was reviewing the data, NEJM still hosted an unblemished PDF of the ADVOCATE study on their website, and Amgen was asserting that the totality of the evidence continued to support approval. Things have changed. The lead authors of the ADVOCATE study, Dr. Peter Merkel and Dr. David Jayne, have now requested retraction, confirming the central fact: endpoint assessments were readjudicated after database lock and unblinding, without their knowledge. The NEJM posted the following:

“The two academic authors of the article by Jayne et al., Avacopan for the Treatment of ANCA-Associated Vasculitis, N Engl J Med 2021;384:599-609,¹ request retraction of the article because, according to an ongoing Food and Drug Administration investigation conducted after publication, and without the knowledge of these two authors, the primary end-point assessments in nine patients were readjudicated after database lock and trial unblinding. This was not disclosed in the article and is inconsistent with proper research conduct. The editors therefore retract the article.”

In short, the core of what the FDA accused ChemoCentryx of doing appears to be true.

Only The Market Always Knows

The reason we discovered this? Aggrieved finance bros.

We only know about this because the 2021 FDA advisory board raised concerns about whether the (already manipulated) efficacy data even supported approval. The advisory board was not aware of the re-adjudication because Bekker and Co did not disclose it. When they split on whether to recommend approval, ChemoCentryx stock collapsed by 80% and furious shareholders sued for securities fraud. In discovery, an expert named Marc Walton produced a report detailing what happened. This prompted the inquiry by the FDA.

That lawsuit was dismissed in August 2025, on summary judgment, without the court ever addressing the manipulation. It was a securities case, which means the question was never whether the trial data were manipulated, it was whether ChemoCentryx had misled investors. As far as I can tell, the court held that the securities-law claims failed. Maddeningly, the alleged misconduct Bekker and Co perpetrated was almost incidental. Had discovery been sealed this may never have come to light.

The one group that seems likely to dodge repercussions is the group of people who are most responsible for this mess. Pirow Bekker and Huibin Yue, for instance, have thus far avoided accountability for regulatory misconduct with multibillion-dollar repercussions. We do not know how much they personally profited from this, but securities plaintiffs alleged that ChemoCentryx CEO Thomas Schall sold roughly $40 million in stock while telling investors the company’s FDA interactions had been “straightforward and routine.” The gulf between “straightforward and routine” and “unprecedented NEJM retraction” is about as wide as you can get.

One of a Kind(ish)

The retraction of any paper is an embarrassing event. It is hard on the scientists who contributed to the paper, whose reputations may be tarnished. It is bad for the journal itself, suggesting inadequacies in their editorial process. It hurts the scientific establishment, which already struggles with a trust-problem.

That does not mean we should avoid retracting questionable studies. If anything, we need more retractions, not fewer. Retractions reflect a healthy scientific process, which is willing to correct itself when it errs. That said, they are objectively bad for everyone involved.

They are also remarkably rare. NEJM has published most of the best medical science since 1812; in the half-century PubMed has tracked it, the journal has logged only 27 retractions over roughly 88,000 papers. That’s an incredible record, something like 0.03%, or one retraction for every 3,300 published articles. It is rarer-still if you count only original research, and it is completely unprecedented if you look at randomized controlled trials that informed regulatory approval. As far as I can tell, this is the first time in the journal’s history that something like this has happened.

Prior retractions have mostly been driven by individual investigators who fabricated data. John Darsee, a Harvard cardiologist, faked data across over 80 papers and abstracts before his 1983 retractions. His hospital became the first institution forced to repay the NIH over research fraud ($122,371), and the scandal helped create the federal Office of Research Integrity. Anil Potti published genomic-signature work that was used in Duke cancer trials but later discredited; those trials were halted, and patients and families sued. Piero Anversa built a career on a cardiac stem cell that probably doesn’t exist, had 31 papers flagged, and cost his institution a $10 million federal settlement.

None of these high profile retractions approach the scope and implications of this one.

The Fallout Snowballs

The FDA led the charge on demanding action, but the momentum is building. The European Medicines Agency recently reached its own verdict, and on June 26, the EMA’s CHMP recommended revoking avacopan’s EU authorization. Like the FDA and now the NEJM, the EMA agrees that ADVOCATE cannot be relied upon. In Europe a CHMP recommendation goes to the European Commission, which typically follows CHMP recommendations.

In January of 2026 when the FDA first requested voluntary withdrawal, most of us did not know the details. That includes myself, most (all?) of my friends in the vasculitis community, and even many (most?) of the people at Amgen. In April when the FDA published the NOOH and laid out their rather shocking case for revoking authorization, I upgraded my concern to “probably will be withdrawn, but we’re not sure if the FDA is correct and perhaps there is more to the story.”

The NEJM retraction changes everything. The study would not have been retracted if the accusations of scientific malfeasance were demonstrably false, and Amgen has yet to refute the accusations leveled by any of the regulatory agencies. I suspect it’s all true. Bekker and Co appear to have cooked the books to cheat the system. The FDA and the EMA and probably the NEJM are furious about this. I am upgrading from “probably will be withdrawn” to “on borrowed time.”

From a medical perspective, I still think avacopan works for ANCA associated vasculitis. I outlined the case for that in my last substack. That said, it is hard to imagine starting any patient on avacopan at this point, and nearly every patient I have explained this to has discontinued it. This is unfortunate, because I do think some patients would benefit from receiving it. It may also be necessary from a public policy perspective. The FDA said it best:

“Withdrawing a drug under these circumstances is critical to protect the public health and the integrity of the drug approval process.”

The Tip of the Iceberg

The biggest question I have after reviewing all of this is a simple one: what if Thomas Schall was being honest? What if manipulating trial data so you can extract billions of dollars from the American public is, in fact, “straightforward and routine?”

That sounds hyperbolic, but consider this: we rely on a system where corporations that stand to make billions of dollars from regulatory approval run the very studies that result in regulatory approval. They fund the institutions that run the studies and they fund the companies that manage the data, help with recruitment, and hire medical writers to write the papers. They also buy reprints, pay open-access or publication charges when applicable, and participate in a publishing economy that benefits from high-profile industry trials.

Remarkably, the system appears to function reasonably well. We have established randomized controlled trials as a (mostly) required step for regulatory approval. Pharmaceutical companies generally run high quality randomized controlled trials, which generally deliver high quality scientific information. This process has brought about hundreds of revolutionary therapies that have saved millions of lives. I know beyond a shadow of doubt that TNF inhibitors work great for rheumatoid arthritis and that rituximab works great for ANCA associated vasculitis, because I see the results every day. Objectively, something about this system works great.

I would also note here that everyone I know in the pharmaceutical industry has been shocked by the details of this. Contrary to what some believe, most of the employees of pharmaceutical companies genuinely want to make the world a better place. Many of them are scientists or doctors who felt like they could have a bigger impact by building new drugs. All of them work in a regulatory environment that causes them to be remarkably cautious, over-indexing on compliance to a degree that would likely be surprising to the lay public.

That slows progress, but it is also good, because all of this requires an enormous amount of trust. We trust that the companies running these studies do not have the capacity to change the outcome of enrolled patients when they see something they do not like. We trust that any protocol deviations will be disclosed at submission and available to the relevant advisory committees. We trust the ostensibly independent trial oversight apparatus to catch deviations like this and report them. The entire regulatory process we have created depends on us placing a remarkable amount of trust in people like Bekker and Yue to behave ethically.