Alzheimer's trial results need more than a statistically significant headline because significance only tells you a difference is unlikely to be chance — not that a patient or family would ever notice it. The two drugs behind the biggest recent headlines, lecanemab and donanemab, both cleared the statistical bar while producing changes on dementia scales that sit well below the level published research says people can actually perceive. That gap matters for anyone weighing treatment for a parent or spouse. The honest picture combines four numbers headlines usually omit: the absolute point difference, the threshold at which a change becomes noticeable, the harm rate, and what the percentage framing actually refers to.
Medical information disclaimer: This article is for general educational purposes only and does not provide medical advice, diagnosis, or treatment. Always consult a physician or other qualified health professional about symptoms, medications, tests, or treatment decisions.
Table of Contents
- What "27% slowing" actually measured
- How big does a change have to be before anyone notices?
- The harms the headlines leave out
- Even the experts disagree about the same numbers
- Questions to bring to the neurologist
- Frequently Asked Questions
What "27% slowing" actually measured
Lecanemab's landmark result came from the phase 3 CLARITY-AD trial of 1,795 people with early Alzheimer's. Over 18 months, the placebo group worsened by 1.66 points on the CDR-SB — an 18-point scale clinicians use to rate dementia severity across memory, judgment, and daily function — while the treated group worsened by 1.21 points. The difference, reported in the New England Journal of Medicine, was 0.45 points, statistically significant at P<0.001. That 0.45-point gap became "27% slowing" because 0.45 is 27% of the placebo group's 1.66-point decline.
The percentage is relative, not absolute. An illustrated analysis in eNeuro notes the same result can be framed as roughly a five-to-six-month delay in decline over the 18-month trial — accurate, but far less dramatic than the headline number. Donanemab followed the same pattern. Its TRAILBLAZER-ALZ 2 trial of 1,736 participants reported "35% slowing" in a favorable subgroup, while the full population's CDR-SB result was about 0.7 points of absolute difference.
How big does a change have to be before anyone notices?
Researchers have measured this directly. The most-cited benchmark, published by Andrews and colleagues in Alzheimer's & Dementia: TRCI, puts the minimal clinically important difference on the CDR-SB at 0.98 points for mild cognitive impairment and 1.63 points for mild Alzheimer's disease.
Set the trial results against that yardstick: This is the core reason a significant p-value is not enough. With nearly 1,800 participants, a trial can reliably detect a difference too small for any individual family to perceive. Statistical power answers "is it real?"; the clinical threshold answers "does it matter?" — and those are different questions.
- Lecanemab: 0.45 points — roughly half the lower threshold, and under a third of the upper one.
- Donanemab: about 0.7 points in the full trial population — still below both thresholds.
- Noticeability benchmark: 0.98 to 1.63 points, depending on disease stage.
The harms the headlines leave out
The significance headline also carries no information about risk, and for anti-amyloid antibodies the risk is concrete. In TRAILBLAZER-ALZ 2, ARIA — amyloid-related imaging abnormalities, meaning brain swelling or effusion visible on MRI — occurred in 24.0% of donanemab patients versus 2.1% on placebo, and three donanemab-treated participants died from serious ARIA, according to the trial report in JAMA. Regulators have read the same data differently for different patients.
The European Commission approved lecanemab in April 2025 only for people who carry zero or one copy of the ApoE4 gene variant, because the EMA's committee judged the ARIA risk in double-carriers too high. In that restricted group, decline slowed 31% with a 9% rate of ARIA-E, per Eisai's announcement. The practical consequence: genetic status, MRI monitoring capacity, and infusion logistics belong in any treatment conversation alongside the efficacy number. A benefit measured in fractions of a point has to be weighed against a harm measured in whole percentage points.
Even the experts disagree about the same numbers
A 2026 Cochrane systematic review pooled 17 randomized trials covering 20,342 participants and concluded that anti-amyloid antibodies produce effects on cognition and dementia severity that are "absent or trivial" — below clinically important thresholds — while raising ARIA risk. Alzheimer's researchers pushed back in a rebuttal published in Alzheimer's & Dementia, arguing the review misapplied individual-level noticeability thresholds to group averages and ignored evidence that benefits grow with longer treatment.
Their point is technical but real: a group mean of 0.45 can contain some patients who benefited substantially and others not at all. For readers, the disagreement is itself the lesson. When credentialed specialists looking at identical significant results reach opposite conclusions about clinical worth, a one-line headline was never going to settle the question.
Questions to bring to the neurologist
Clinicians writing in Alzheimer's Research & Therapy argue trial reports should present absolute point differences, individual variability, and harms alongside the p-value — and families can apply the same standard in the clinic. For these drugs, the honest group-average summary is a delay of a few months in decline over 18 months, an ongoing infusion schedule with MRI safety monitoring, and a real ARIA risk. Useful questions to ask: The European restriction is a working example of this arithmetic: the same drug, judged worth the risk for one genetic group and not another, based on the same trial.
- What is the absolute expected benefit on the scale used, not the relative percentage?
- Has the patient been tested for ApoE4 status, and how does the result change the risk?
- What does the MRI monitoring schedule involve, and how quickly would ARIA be caught?
- Given this patient's stage, does the expected benefit reach the level anyone would notice?
Frequently Asked Questions
Does "statistically significant" mean the drug works well?
No. It means the measured difference is unlikely to be chance. Lecanemab's 0.45-point CDR-SB benefit was significant at P<0.001 yet sits below the 0.98–1.63 point range research identifies as the smallest change patients and clinicians notice.
Why do trials report percentages instead of points?
The percentage compares the treated group's decline with placebo's, so "27% slowing" means 0.45 fewer points of decline against placebo's 1.66. Relative framing is accurate but reads as far larger than the absolute difference.
Is ARIA common on these drugs?
In donanemab's phase 3 trial, brain swelling or effusion occurred in 24% of treated patients versus 2.1% on placebo, and three treated participants died from serious ARIA, which is why MRI monitoring is required.
