EMA Revokes Avacopan (Tavneos) After GCP Breaches in ADVOCATE Trial, as FDA Hearing Looms
核心洞察
The European Commission revoked avacopan (Tavneos) across the EU and EEA on August 6, 2026, after the CHMP found "serious breaches" of Good Clinical Practice in the pivotal Phase 3 ADVOCATE trial.
Nine patient records were re-adjudicated after database lock and unblinding, with five avacopan-arm patients changed to "sustained remission," converting a non-superior week-52 result into a superiority finding.
The EMA ruled that a company-commissioned Duke Clinical Research Institute re-adjudication cannot restore GCP compliance, even though it confirmed noninferiority but failed to reproduce the original superiority claim.
The European Medicines Agency (EMA) has published the full scientific rationale behind Europe's withdrawal of avacopan (Tavneos), documenting how nine altered patient records—changed after a trial's database was sealed and treatment assignments were revealed—destroyed the evidentiary foundation of a $459 million drug approval. The EMA's Article 20 Public Assessment Report, published August 6, 2026, explains the grounds for the European Commission's formal revocation of avacopan across all EU member states and the European Economic Area, following a CHMP recommendation to revoke adopted on June 25, 2026.
For patients and clinicians in the United States, where Tavneos remains FDA-approved and available while the agency evaluates Amgen's July 23 formal hearing submission, the EMA's published reasoning carries direct implications. The same data-integrity failure that ended avacopan's four-year European run is the basis for the FDA's own proposal to withdraw Tavneos, issued April 27, 2026.
What "Database Lock" Means and Why Breaking It Matters
In a blinded randomized trial, a database lock is the point at which data collection is formally declared complete. Access to the data is sealed, no further changes are permitted without a formal documented amendment, and unblinding occurs only after the lock. This sequence exists because if anyone with knowledge of which patients received the experimental drug can alter patient outcome assessments, the results of the trial can be biased. This requirement is codified in the ICH E6 Good Clinical Practice guidelines.
In the ADVOCATE trial, according to the EMA's published assessment report and the FDA's April 27 NOOH letter, that sequence was violated in a specific and consequential way.
What the EMA Found Happened in November 2019
The ADVOCATE trial enrolled 331 patients with granulomatosis with polyangiitis (搜索) (GPA) or microscopic polyangiitis (搜索) (MPA)—the two main forms of ANCA-associated vasculitis (搜索), rare autoimmune conditions in which the immune system attacks small blood vessels, causing organ damage that can be fatal without aggressive treatment. The trial (NCT02994927) compared avacopan against a standard high-dose corticosteroid taper, both combined with either rituximab or cyclophosphamide.
Avacopan works by blocking the C5a receptor (搜索) (C5aR1 (搜索)) on neutrophils, disrupting a self-amplifying inflammatory loop central to ANCA vasculitis pathology. Blocking C5aR1 with avacopan interrupts this loop while theoretically allowing corticosteroid dose reduction.
According to the EMA's published assessment report, after the first database lock on November 5, 2019, sponsor personnel with access to the unblinded data identified that the ADVOCATE trial had not met one of its most important findings: superiority at week 52 for sustained remission. Nine patients were then selected for re-adjudication of their primary endpoint assessments. At week 52, five of those nine patients changed status. All five were in the avacopan group; all five moved from "not in sustained remission" to "sustained remission."
The FDA's investigation identified two ChemoCentryx (搜索) employees who had access to unblinded data and were involved in this process: Huibin Yue, PhD, the company's director of biostatistics, and Pirow Bekker, MD, PhD, who served as ChemoCentryx's chief medical officer and is listed as a co-author of the original ADVOCATE trial paper published in the New England Journal of Medicine.
None of this appeared in the NDA submission or in the 2021 NEJM publication. The academic leaders of the trial—Dr. David Jayne of Cambridge and Dr. Peter Merkel of the University of Pennsylvania—were unaware it had occurred. When they learned of it through the FDA investigation, they requested the retraction. The NEJM retraction notice, published June 29, 2026, describes the conduct as "inconsistent with proper research conduct."
Why Amgen's Independent Re-Analysis Couldn't Save the Drug in Europe
After the data manipulation came to light, Amgen commissioned the Duke Clinical Research Institute (DCRI) to conduct a fully blinded independent re-adjudication of the ADVOCATE primary endpoints. The DCRI re-adjudication confirmed noninferiority of avacopan versus the prednisone taper for remission at week 26—68.1% versus 67.1%, a difference of 2.2 percentage points—and for sustained remission at week 52, 61.4% versus 52.4%, a 9.8 percentage point difference.
However, the DCRI re-adjudication did not reproduce the original superiority finding at week 52. The original ADVOCATE analysis had reported a 12.5 percentage point superiority (95% CI, 2.6% to 22.3%)—a statistically significant result. The DCRI re-adjudication produced a 9.8 percentage point difference (95% CI, -0.3 to 19.9)—not statistically significant.
The CHMP considered the DCRI re-adjudication and rejected it as a remedy. Once a clinical trial's database has been locked and treatment assignments unblinded, any subsequent re-adjudication of primary endpoints—even by an entirely independent external party—cannot restore GCP compliance to the original data, because the contamination of the data environment is structural and irreversible. The CHMP's conclusion, adopted June 25, 2026, was that the ADVOCATE study was conducted in breach of Good Clinical Practice principles, the data cannot be relied upon to demonstrate Tavneos's effectiveness, and post-marketing data and post-hoc analyses are insufficient to fill that evidentiary gap.
A Parallel Safety Signal and Its Independent Weight
While the GCP violation is the primary driver of regulatory action, a parallel pharmacovigilance crisis has deepened the case against avacopan. In its March 31, 2026 drug safety communication, the FDA identified 76 cases of drug-induced liver injury (搜索) (DILI) with reasonable evidence of causal association with avacopan—74 serious outcomes, 54 hospitalizations, and 8 deaths in the global dataset.
The most acute picture is in Japan. Kissei Pharmaceutical (搜索) disclosed in May 2026 that 20 patients had died after taking Tavneos in Japan since the drug went on sale in 2022, including 13 deaths among 22 cases of vanishing bile duct syndrome (搜索). Amgen has noted that as of its July 23 submission, no confirmed VBDS deaths have been attributed to Tavneos among the more than 8,000 patients treated in the United States. The company updated US prescribing information on May 29, 2026, adding stronger liver monitoring requirements.
The CHMP's revocation rationale explicitly relied on both the data integrity failure and the hepatotoxicity risk: with the sole pivotal trial unusable and post-marketing evidence insufficient to replace it, the serious liver risks are no longer outweighed by a proven benefit.
A Precedent for Every Drug Approved on a Single Pivotal Trial
Most major drug approvals rest on at least two adequate and well-controlled trials. The FDA and EMA occasionally permit approval on the basis of a single pivotal study—as they did with avacopan, in part because ANCA-associated vasculitis (搜索) is rare—but this pathway leaves the approval more exposed if that single study is subsequently compromised.
The EMA has now established that a post-approval Good Clinical Practice finding that invalidates the sole pivotal study is sufficient grounds for EU revocation—without requiring a concurrent catastrophic new safety signal. A drug can be pulled from the European market solely because the data that justified its approval can no longer be trusted, even if post-marketing real-world evidence and a company-commissioned independent re-analysis suggest it probably works.
Amgen Contests the US Withdrawal
In the United States, the story is not over. Tavneos remains FDA-approved and commercially available, with US sales of $119 million in the first quarter of 2026 alone—a 32% year-over-year increase—according to Amgen's Q1 2026 financial results.
On July 23, 2026, Amgen formally submitted its package in support of a public hearing, contesting the FDA's April 27 withdrawal proposal. The submission includes the DCRI re-adjudication results, a systematic literature review and meta-analysis covering 71 real-world studies involving more than 2,200 patients (remission rates of 87% at six months and 93% at 12 months), post-marketing safety data, and patient and clinician testimonials.
Amgen's legal position is that the FDA's withdrawal proposal rests on a disputed characterization of the 2019 re-adjudication process, that the ADVOCATE trial as a whole still constitutes adequate and well-controlled evidence of effectiveness, and that the DCRI's fully blinded independent re-adjudication represents a sufficient remediation. The CHMP considered essentially the same argument and rejected it. Whether the FDA takes the same view is the regulatory question that will determine whether avacopan survives in any major market.
The patient community's position was stated clearly by Vasculitis International after the CHMP recommendation: avacopan addressed "a specific and underappreciated burden"—the impact of glucocorticosteroids on patients' daily lives—and removing it leaves clinicians without a tool specifically designed to reduce that burden.