The first drug proven to slow Alzheimer's

Investigators of the Clarity AD trial showed in 2023 that lecanemab slowed cognitive decline by 27 percent over 18 months, the first drug to change the course of Alzheimer's disease.

A pipeline that produced almost nothing

For most of the modern era of Alzheimer's research, the drugs that reached patients treated symptoms and left the disease untouched. In 2014 Cummings and colleagues counted what the field had tried. Between 2002 and 2012, 413 Alzheimer's trials were registered on ClinicalTrials.gov: 124 in phase 1, 206 in phase 2 and 83 in phase 3. Across that decade the overall success rate was 0.4 percent, a failure rate of 99.6 percent, which the authors described as among the lowest in any therapeutic area. Five drugs were approved at the time, four cholinesterase inhibitors and one NMDA receptor antagonist. None of them slowed the underlying disease.

The target most of those failed trials aimed at was amyloid-beta. In 1992 Hardy and Higgins set out in Science what became known as the amyloid cascade hypothesis: that the accumulation of amyloid-beta in the brain is the initiating event in Alzheimer's disease, with tangles, cell loss and dementia following from it. For three decades the hypothesis was tested by trying to remove amyloid and watching whether cognition changed. Until Clarity AD, it had not.

An antibody aimed at protofibrils

Lecanemab is a humanized IgG1 monoclonal antibody. It was engineered to bind with high affinity to soluble amyloid-beta protofibrils, the intermediate aggregates that form before insoluble plaques, rather than to single amyloid molecules or to mature plaque alone. The reasoning, as the Clarity AD investigators put it in the New England Journal of Medicine, was that both soluble and insoluble aggregated amyloid "may initiate or potentiate pathologic processes" in the disease.

A phase 2 trial that missed its own target

The antibody's first large test was Study 201, reported by Swanson and colleagues in 2021. It enrolled 854 people with early Alzheimer's disease, 609 assigned to lecanemab across several dose regimens and 245 to placebo, in a Bayesian design that shifted later participants toward the doses that appeared to be working. The primary endpoint was set in advance: an 80 percent probability that the best dose reduced 12-month clinical decline by at least 25 percent on a composite scale called ADCOMS. The trial reached a probability of 64 percent. It missed.

The prespecified 18-month analyses told a different story. At 10 milligrams per kilogram every two weeks, the dose carried forward, brain amyloid fell by 0.306 standardized uptake value ratio units, and decline on the Clinical Dementia Rating Sum of Boxes was 33 percent less than placebo by the Bayesian analysis and 26 percent less by the frequentist one. Amyloid-related imaging abnormalities with edema occurred in 9.9 percent of participants at that dose. The authors wrote that a phase 3 trial, Clarity AD, was under way.

Clarity AD: 1,795 people over 18 months

van Dyck and colleagues published Clarity AD in the New England Journal of Medicine in 2023. It was a multicenter, double-blind, phase 3 trial of adults aged 50 to 90 with mild cognitive impairment or mild dementia due to Alzheimer's disease and amyloid confirmed by PET scan or cerebrospinal fluid. A total of 1,795 participants were randomized one to one, 898 to intravenous lecanemab at 10 milligrams per kilogram every two weeks and 897 to placebo, for 18 months.

The primary endpoint was the change in the Clinical Dementia Rating Sum of Boxes, or CDR-SB, a clinician-rated scale from 0 to 18 that combines cognition and daily function, with higher scores meaning greater impairment. Both groups began at a mean of about 3.2. At 18 months the adjusted mean change was 1.21 points with lecanemab and 1.66 with placebo, a difference of 0.45 points with a 95 percent confidence interval of 0.23 to 0.67 and a P value below 0.001.

The 27 percent figure in this story is the ratio of those two numbers: a 0.45-point difference set against the 1.66 points of decline in the placebo group. It is a relative measure. In absolute terms, the treated group declined by about three quarters of what the untreated group declined over a year and a half, on an 18-point scale.

The secondary endpoints moved in the same direction. In a substudy of 698 participants, amyloid on PET fell by 59.1 centiloids more with lecanemab than with placebo. On the 14-item cognitive subscale of the Alzheimer's Disease Assessment Scale, the difference favoring lecanemab was 1.44 points on a 90-point scale. On the activities-of-daily-living scale for mild cognitive impairment, it was 2.0 points on a 53-point scale. The authors summarized the result as "moderately less decline on measures of cognition and function than placebo at 18 months," and added that longer trials were warranted.

What the drug cost the people who took it

Infusion-related reactions occurred in 26.4 percent of lecanemab recipients. Amyloid-related imaging abnormalities with edema or effusions, known as ARIA-E, occurred in 12.6 percent. These are swelling or fluid in the brain visible on MRI, most often without symptoms, but not always. When the FDA converted the drug to traditional approval it attached a boxed warning stating that ARIA can "infrequently cause serious, life-threatening brain edema associated with seizures," that intracerebral hemorrhages "can occur and be fatal," and that people carrying two copies of the ApoE ε4 gene have a higher incidence of ARIA, so that testing for the gene should be done before treatment begins.

The trial that came before, and why one clean result mattered

Lecanemab was not the first anti-amyloid antibody to reach regulators. Aducanumab, tested in the twin phase 3 trials EMERGE and ENGAGE with 1,638 and 1,647 participants, was halted for futility after an interim analysis of roughly the first half of enrolled patients. When the full data were analyzed, EMERGE met its primary endpoint with a 0.39-point CDR-SB difference at the high dose, a 22 percent reduction in decline, while ENGAGE showed a 0.03-point difference in the wrong direction. Two trials of the same drug disagreed, and the field was left arguing about which one to believe.

Clarity AD was a single trial that met its primary endpoint and every key secondary endpoint, in the direction predicted, with a confidence interval that excluded zero. That is the specific sense in which the curve moved.

From accelerated to traditional approval

The FDA's approval letter for Leqembi, the brand name of lecanemab, was signed on January 6, 2023, under the accelerated approval pathway, which rests on a surrogate endpoint, in this case the reduction of amyloid plaques, rather than proven clinical benefit. On July 6, 2023 the agency converted that to a traditional approval on the basis of Clarity AD. Teresa Buracchio, then acting director of the FDA's Office of Neuroscience, described the action as "the first verification that a drug targeting the underlying disease process of Alzheimer's disease has shown clinical benefit."

What the story does and does not show

Clarity AD showed, in 1,795 people with early-stage disease and confirmed amyloid, that 18 months of lecanemab produced 0.45 points less decline on an 18-point scale than placebo. It did not show improvement; both groups got worse. It did not test people with moderate or severe dementia, or people without amyloid, and its result cannot be extended to them. Eighteen months is the full length of the evidence, and whether the gap between the curves widens, holds or closes over five or ten years is unknown. The trial did not measure whether a 0.45-point difference is one that patients and families notice; that question was argued afterward, not settled by the data.

The trial did show that removing amyloid from the brain changed a clinical outcome in a randomized, placebo-controlled test, after three decades in which that had never happened. It also showed that the change was modest and the adverse events were real. Both halves of that sentence are in the primary paper, and this story reports both.

After a century of failure, the curve moved.