flowchart TD
A[Elderly patient<br/>Vascular fragility] --> B[Anticoagulants<br/>/ Antiplatelets]
A --> C[Lecanemab]
C --> D[Amyloid removal]
D --> H[Modest slowing<br/>of decline]
D --> E[ARIA-E / ARIA-H]
B --> E
E --> F[Edema / Microbleeds / ICH]
F --> G[Hospitalization<br/>Functional decline]
H -. small benefit vs risk .- G
Lecanemab in Early Alzheimer’s Disease
Medicine
The paper is a bit old, but it’s interesting; I learn from reading and writing about it so here’s my attempt.
The lecanemab paper reads, at first glance, like the trial we’ve been waiting for (Dyck et al. 2023).
This is a big phase 3 trial in NEJM where an anti-amyloid antibody finally does what the field has been promising for quite some time. It strips amyloid out of the brain and the treatment group declines more slowly than placebo over eighteen months. It’s statistically clean, biomarkers all pointing in the right direction, figures that slope apart in the way grant proposals like to appeal.
Technically you are removing material; structurally it might be too late.
The core biology is pretty familiar. Amyloid-β is shaved off the amyloid precursor protein, misfolds, and aggregates into extracellular plaques. Tau, inside the neuron, becomes hyperphosphorylated and tangles up the microtubules that were supposed to carry cargo; synapses thin out; networks lose coherence (Kumar et al. 2021). In parallel, the vasculature of an aging brain quietly turns hostile: small-vessel disease, and crucially, CAA not in the parenchyma this time, but packed into the walls of tiny arteries and arterioles.
Lecanemab is a monoclonal antibody against soluble amyloid protofibrils, given as an intravenous infusion every two weeks. In this trial, about 1,800 people with “early” Alzheimer’s—mild cognitive impairment or mild dementia, all amyloid-positive on PET or CSF—were randomized to drug or placebo and followed for eighteen months. The primary outcome was change in the Clinical Dementia Rating Sum of Boxes (CDR-SB), a clinician-rated 0–18 scale that tries to compress memory, orientation, judgment, community affairs, hobbies, and personal care into a single trajectory.
On that metric, the placebo group worsened by about 1.66 points. The lecanemab group worsened by about 1.21. The difference—roughly 0.45 CDR-SB points (95% confidence interval, −0.67 to −0.23; P<0.001)—is what becomes the “27% slowing of decline”. Amyloid PET scans show the drug is doing what it says: plaque signal falls dramatically compared with placebo. Fluid biomarkers like phosphorylated tau drift in the right direction. At the level of graphs and p-values, the paper is satisfyingly coherent.
But you have to translate these results back into an actual elderly human being.
Real patients are not the stylized creatures of trial diagrams. They come with AFib and a history of deep vein thrombosis and coronary stents and that one “small stroke” no one talks about at family dinners. They are already on warfarin or a direct oral anticoagulant to keep clots from forming in AFib; or on aspirin and clopidogrel to keep a metal stent in the left anterior descending from occluding; or both. These drugs flatten the normal peaks of coagulation so that clots are slower to form, bleeds are harder to stop, and any damage, even slight, to a vessel wall has more room to become a crisis.
The new class of anti-amyloid antibodies all carry the side effect, ARIA. ARIA-E is vasogenic edema or effusion, pockets of swelling or fluid that show in white on the MRI. ARIA-H is microhemorrhages that leave scars. In this trial, roughly a quarter of patients had infusion reactions, and around twelve to thirteen percent developed ARIA-E; ARIA-H also increased in the active arm. Most of these findings were asymptomatic and discovered on scheduled scans, but not all. Some patients had headache, confusion, or seizures; post-marketing reports have documented rare but real fatal hemorrhages. The FDA label now comes with a boxed warning spelling out the risks, especially in people on anticoagulants.
None of this is surprising. Anti-amyloid antibodies strip amyloid off vessel walls as well as off plaques. If those vessels are already stiff, hypertensive, and studded with amyloid from decades of cerebral amyloid angiopathy, “cleaning” them can mean destabilizing them. The wall leaks a little: ARIA-E. The wall cracks a little: ARIA-H. Add systemic anticoagulation or aggressive antiplatelet therapy to that fragility and the margin between “tiny microbleed that never matters” and “lobar hemorrhage that changes the course of your life” gets uncomfortably thin.
The trial tries to manage this with MRIs: baseline imaging to exclude anyone with too many microbleeds and follow-up scans at defined time points. This schedule alone is a selection pressure: to even begin, you need a body that tolerates lying still in a noisy tube at these points in time. Once you move from the idealized trial environment to a real clinic that already can’t get outpatients into MRI within a month, the whole premise is undermined.
Meanwhile the patient is supposed to show up every two weeks, indefinitely, for an intravenous infusion. For someone in their forties, this is boring. For someone in their eighties who walks with a frame and gets disoriented on unfamiliar floors, every visit is an opportunity for a fall, a urinary infection, a bout of hospital delirium that sets them back more than any fraction of a CDR-SB point can capture (remember that this is not a perfect metric).
There is also the quieter question of what “0.45 points” buys you, existentially.
The paper, reads like a proof of concept for a pathophysiological story, not a compelling offer for the average person who actually has this disease. It tells us that amyloid matters, that if you remove a lot of it early enough you can nudge the clinical trajectory in the right direction. It also tells us, indirectly, that by the time the diagnosis is even on the table, the brain is already woven through with other forms of damage that an antibody cannot touch.
References
Dyck, Christopher H. van, Chad J. Swanson, Paul Aisen, Randall J. Bateman, Christopher Chen, Michelle Gee, Michio Kanekiyo, et al. 2023. “Lecanemab in Early Alzheimer’s Disease.” New England Journal of Medicine 388 (1): 9–21. https://doi.org/10.1056/NEJMoa2212948.
Kumar, Vinay, Abul K. Abbas, Jon C. Aster, Jerrold R. Turner, James Alfred Perkins, Stanley L. Robbins, and Ramzi S. Cotran, eds. 2021. Robbins & Cotran Pathologic Basis of Disease. Tenth edition. Philadelphia, PA: Elsevier.