ARIA, short for amyloid-related imaging abnormalities, is the brain swelling or small bleeding that can appear on MRI in some people who take the newer anti-amyloid antibodies for early Alzheimer's disease, lecanemab and donanemab. It shows up in two patterns: ARIA-E, meaning fluid or edema, and ARIA-H, meaning tiny hemorrhages or iron staining. Most cases occur in the first months of treatment and cause no symptoms, though a small minority are serious and rare deaths have been reported. The practical news for patients and clinicians is that dosing, genetic testing, and scan timing have all been revised, and each change lowers the risk in a way that can be measured.
Key points#
- ARIA comes in two forms: ARIA-E (swelling) and ARIA-H (microbleeds or iron deposits), and most of it is found on routine scans rather than because a person feels unwell.
- Carrying two copies of the APOE4 gene is the single strongest predictor of symptomatic ARIA, so testing before treatment helps quantify the risk.
- A gentler donanemab dose ramp cut ARIA-E substantially in a randomized trial while still clearing amyloid, and the FDA updated the label to match.
- For lecanemab, an extra earlier MRI was added so problems are caught before they cause symptoms.
Why the scans matter more than the symptoms#
It helps to think of ARIA as something the MRI sees before the person feels it. When symptoms do occur, they tend to be nonspecific: headache, confusion, dizziness, or changes in vision. Because these overlap with ordinary complaints and because most ARIA is silent, the imaging schedule, not how you feel on any given day, is what actually protects people. That is the reasoning behind fixed monitoring scans at set points during the first year of treatment.
The scale is worth stating plainly. In the pivotal lecanemab trial, roughly 12.6 percent of treated participants developed ARIA-E and 17.3 percent developed ARIA-H, as summarized in a review in Brain Communications. For donanemab across its phase 3 program, ARIA-E ran near 24 percent and ARIA-H near 31 percent. The great majority of these findings were picked up on scheduled imaging in people who felt fine.
What is happening inside the vessels#
These antibodies bind amyloid plaque so the immune system can clear it. That clearance appears to put temporary stress on small blood vessels, particularly vessels whose walls are already loaded with amyloid, a condition known as cerebral amyloid angiopathy. As amyloid is pulled out, the vessel can leak fluid (ARIA-E) or shed small amounts of blood and iron (ARIA-H). The Brain Communications review describes autopsy findings in one lecanemab trial participant as severely damaged vessel walls with reactive immune cells, which fits the picture of blood vessels that are briefly weakened while they are being cleared.
Genetics sets the baseline odds#
One gene, APOE, shifts the whole risk picture. People who carry two copies of the APOE4 variant, called homozygotes, sit at the high end. In the lecanemab data, symptomatic ARIA-E occurred in about 9.2 percent of APOE4 homozygotes, compared with roughly 1.7 percent of those with a single copy and 1.4 percent of non-carriers, and severe imaging findings followed the same steep gradient. For that reason the current FDA label advises checking APOE4 status before the first dose. The point is not to rule anyone out automatically, but to let you and your clinician see the true size of the risk before deciding. Two other baseline features raise risk in a comparable way, namely already having several small hemorrhages visible on MRI and taking blood thinners. The review notes that two or more microhemorrhages at baseline roughly doubled the chance of developing ARIA-E, which is why a baseline scan and a careful look at your current medications belong before treatment starts, not after.
A gentler dose ramp, tested head to head#
The clearest new evidence concerns how donanemab is started. Rather than moving quickly to the full dose, a modified schedule spreads the early doses out so the ramp-up is slower. In the TRAILBLAZER-ALZ 6 study, published in 2025 in a peer-reviewed Alzheimer's journal, this modified titration reduced ARIA-E from about 23.7 percent to 13.7 percent at 24 weeks, a relative drop of roughly 42 percent, and from about 24.2 percent to 15.6 percent at 52 weeks, roughly 35 percent. Amyloid was still cleared to a similar degree, with adjusted mean plaque changes of around 71 and 72 centiloids across the two schedules. In everyday terms, starting slower kept most of the benefit while removing a meaningful share of the swelling.
On the strength of that result, the FDA approved an updated Kisunla (donanemab) label in 2025 that adopts the modified titration for early symptomatic Alzheimer's disease, per Eli Lilly's announcement of the approval. It is a useful reminder that a change in how a drug is dosed, not a new molecule, can move safety.
Earlier eyes on lecanemab#
For lecanemab, the adjustment was to the scan schedule rather than the dose. After reviewing serious cases, the FDA issued a drug safety communication recommending an additional, earlier monitoring MRI before the third infusion, on top of the scans already advised before the fifth, seventh, and fourteenth infusions. The logic mirrors the biology: because ARIA clusters early, looking sooner can catch it before it becomes symptomatic, at which point treatment is usually paused until the imaging settles. The review literature also stresses scan quality, favoring higher-field MRI with sequences that are sensitive to small bleeds so findings are not missed.
How to weigh it as a patient#
The honest summary is that these therapies remove amyloid and modestly slow decline, and that they carry a real but mostly manageable imaging risk that is far better mapped now than it was at first approval. None of the recent changes make ARIA disappear. What they do is sort out who is most vulnerable, start higher-risk regimens more gently, and look earlier. If you are considering treatment, the concrete questions are the same short list: your APOE4 status, what your baseline MRI shows, whether blood thinners are in the picture, and whether the center can perform and read the monitoring scans on schedule. Together those factors describe the actual risk far better than any single headline percentage.
Sources and further reading
Questions and answers
Is ARIA usually dangerous?
Most ARIA is found on routine MRI and causes no symptoms, and it often settles once treatment is paused. A minority of cases are serious, however, which is why baseline testing and scheduled scans are built into treatment.
Should everyone get APOE4 testing before starting?
Current FDA labeling advises checking APOE4 status before the first dose, because carrying two copies substantially raises the risk of symptomatic ARIA. The result informs the decision rather than automatically ruling treatment in or out.
Does slower dosing mean the drug works less well?
In the donanemab titration trial, the gentler schedule cleared amyloid to a similar degree as the faster one while producing less ARIA-E, so the trade appeared to preserve benefit while improving safety.