Anti-Amyloid Infusion Therapy for Early Alzheimer’s Disease

Alzheimer’s treatment safety guide

A detailed review of randomized and long-term benefit, real-world evidence, ARIA risk, MRI monitoring, APOE testing, infusion reactions, and emergency precautions.

Anton Ostashko, MD·Last medically reviewed July 12, 2026

Overview

Leqembi® (lecanemab) and Kisunla™ (donanemab) are anti-amyloid monoclonal antibodies for selected people with mild cognitive impairment or mild dementia due to Alzheimer’s disease and confirmed brain amyloid. They remove amyloid plaques and can modestly slow the average rate of cognitive and functional decline. These are disease-modifying treatments: their purpose is to slow how quickly Alzheimer’s disease worsens, not to make memory noticeably better right away. They are not cures, do not stop the disease, and do not restore abilities already lost.1234

The most important treatment risk is amyloid-related imaging abnormalities (ARIA): brain swelling or fluid called ARIA-E, and small areas of bleeding or iron deposition called ARIA-H. ARIA is often found only on a scheduled MRI, but it can cause symptoms, seizures, stroke-like deficits, permanent disability, or death. Both current U.S. labels carry boxed warnings for ARIA.12

Who was studiedEarly symptomatic AD

Mild cognitive impairment or mild dementia due to Alzheimer’s disease, with confirmed amyloid.

Benefit shownSlowing that can accumulate

Randomized trials showed modest slowing by about 18 months; extension analyses show continued separation through 3 to 4 years.

Main safety issueARIA on MRI

Often asymptomatic, but seizures, stroke-like deficits, disability, hemorrhage, and death can occur.

What treatment may do

  • Lower the average rate of cognitive and functional decline.
  • Keep patients in an earlier disease stage longer, on average.
  • Build a larger group-average time gap when slowing persists over several years.
  • Reduce amyloid plaque in the brain.

What treatment does not do

  • It does not cure Alzheimer’s disease or stop all decline.
  • It does not restore abilities already lost.
  • It does not guarantee that one person will gain a stated number of months.
  • It does not make extension or real-world comparisons equivalent to a continuously randomized placebo trial.

Important context: Leqembi and Kisunla were studied in different trials, with different dosing, sample sizes, MRI schedules, and follow-up. The percentages below are useful for informed discussion, but they are not a head-to-head comparison and cannot predict one person’s outcome.

Evidence cutoff: This guide reflects publicly available evidence and current U.S. prescribing information through July 12, 2026. The main Alzheimer’s Association International Conference begins July 12, 2026; results scheduled for presentation after this cutoff are not treated here as established evidence.

PDF

Printable patient and caregiver reference

A concise handout that summarizes expected benefits, treatment and MRI schedules, overall ARIA risk and by APOE genotype, medication precautions, and symptoms that require urgent care.

Download the Leqembi and Kisunla Patient Guide (PDF) General educational reference: this handout does not replace individualized medical advice or the informed-consent discussion with the treating clinician.

Who May Be Considered for Treatment?

Treatment is generally considered only when all of the following are true:

  • Symptoms are at the mild cognitive impairment or mild dementia stage, and Alzheimer’s disease is the most likely primary cause.
  • Amyloid pathology has been confirmed with an accepted test, commonly amyloid PET or cerebrospinal fluid testing.
  • A recent brain MRI does not show a bleeding pattern, severe vascular injury, or another finding that makes treatment unacceptably risky.
  • The patient and care partner can complete infusions or injections, scheduled MRIs, symptom monitoring, and follow-up visits.
  • The expected benefit, ARIA risk, treatment burden, cost, and the patient’s goals support treatment.

These medicines have not been shown to help when started in people without cognitive symptoms, in moderate or severe dementia, or when another disease is the main cause of impairment.12


How Leqembi and Kisunla Differ

Both medicines remove amyloid, but they bind different forms of amyloid and use different treatment strategies. These distinctions help explain the dosing plans; they do not prove that one medicine is clinically superior for a particular patient.122021

FeatureLeqembi (lecanemab)Kisunla (donanemab)
Primary binding profileAggregated soluble and insoluble amyloid-beta, with particular affinity for soluble protofibril assemblies.N-terminal pyroglutamate-modified amyloid-beta concentrated in deposited plaques.
Clinical strategyOngoing treatment: an 18-month initiation phase followed by an approved maintenance option.Plaque-clearance strategy: treatment may be stopped after PET shows minimal amyloid, using individualized clinical judgment.
What the distinction does not establishTarget biology alone cannot tell an individual patient which medicine will produce greater clinical benefit or lower overall risk. There has been no direct head-to-head efficacy trial.

How the targets differ in plain language: Amyloid exists in several forms. Leqembi binds aggregated soluble amyloid, including protofibrils, as well as insoluble amyloid in plaques. Kisunla binds a modified insoluble form of amyloid found mainly in established plaques. Both reduce amyloid plaque on PET scans. Neither directly targets tau tangles, and no head-to-head trial has shown that these target differences make one medicine more effective or safer for a particular patient.1234

What Benefit Has Been Demonstrated?

The evidence comes from randomized trials, longer-term extension studies, and experience in clinical practice. These sources answer different questions, and the randomized trials remain the strongest evidence that the medicines caused the observed difference in decline.

Randomized trialsAbout 18 months

Randomized, placebo-controlled trials show that both medicines slow average cognitive and functional decline.

Longer-term follow-up3 to 4 years

Longer follow-up suggests that the benefit continues, but there was no ongoing randomized placebo group, so these estimates are less certain.

Clinical-practice dataFeasibility and safety

Registries and claims data help show how treatment is delivered and continued; they are less reliable for proving effectiveness.

Results from the randomized trials

Both pivotal trials used the Clinical Dementia Rating-Sum of Boxes (CDR-SB), an 18-point measure combining cognition and daily function. Higher scores mean greater impairment.

Medicine and trialAverage worsening with treatmentAverage worsening with placeboAbsolute differenceRelative slowing
Leqembi, 18 months1.21 points1.66 points0.45 point less worsening27%
Kisunla, 76 weeks1.72 points2.42 points0.70 point less worseningAbout 29%

These percentages do not mean that patients improved by 27% or 29%. Both treated and placebo groups worsened; the treated groups worsened somewhat less on average.34

Putting the difference into time

“Time saved” does not mean extra life expectancy. It means that, on average, the treated group reached a similar level of measured worsening later than the comparison group. During the randomized trial periods, this worked out to roughly 5 to 6 months of time-equivalent slowing by about 18 months with either medicine.1314

Leqembi long-term estimateAbout 12 months at 4 years

A dedicated time-component analysis estimated 11.0 months versus ADNI and 13.1 months versus BioFINDER.13

Kisunla long-term estimateAbout 7 months at 3 years

The peer-reviewed three-year follow-up estimated about seven months of average time-equivalent slowing compared with a matched untreated group.10

These estimates describe the average difference between groups. They do not mean extra life expectancy, a symptom-free interval, or a guaranteed delay for an individual. Some people decline faster, some more slowly, and some may have no detectable benefit.

Kisunla: what the 36-month data add

The peer-reviewed TRAILBLAZER-ALZ 2 follow-up tracked participants for about three years. Compared with a matched untreated group from the Alzheimer’s Disease Neuroimaging Initiative (ADNI), participants who started donanemab earlier had 1.2 points less CDR-SB worsening at three years (95% confidence interval, 0.7 to 1.7 points less worsening). Starting earlier was also associated with a 27% lower risk of progressing to the next CDR-Global stage than starting after the 76-week placebo period. More than 75% of assessed participants reached amyloid clearance, and benefit remained evident in people who had completed the treatment course.10

About seven months at three years

At three years, the early-start group reached a similar level of cognitive and functional decline about seven months later than the matched comparison group. This is an average estimate, not a guaranteed delay for one person.10

Kisunla’s PET-Based Treatment-Completion Strategy

In the pivotal program, 47% (313 of 672) of treated participants met the trial’s PET-based completion criteria by 12 months, and 69% (429 of 620) met them by 18 months. Completion meant an amyloid level below 11 Centiloids on one scan, or 11 to less than 25 Centiloids on two consecutive scans. It did not mean that Alzheimer’s disease had been cured or that cognition had stopped changing.216

The current Lilly HCP summary reports that 85% of participants in the long-term extension met the trial’s PET completion criteria within three years. The peer-reviewed extension reports amyloid clearance in more than 75% of participants who had an evaluable scan; the percentages are not identical because the analytic populations and denominators differ.1016

In a smaller subgroup of 323 participants who met the PET completion criteria by one year, the 36-month CDR-SB difference was 1.3 points in favor of early donanemab compared with matched ADNI data. This suggests that benefit may persist after active dosing ends, but the analysis had no continuing placebo group and is less certain than the original randomized trial.16

Amyloid can return after treatment stops. The long-term report estimated relatively slow reaccumulation—approximately 2.4 Centiloids per year—and the current label reports a similar median estimate of 2.80 Centiloids per year. There is not yet a validated universal schedule for repeat PET imaging, retreatment, or restarting Kisunla after amyloid reaccumulates.210

Treatment completion does not mean cure. It is a PET-defined dosing milestone, not proof that the disease is gone, that the patient will remain in MCI, or that treatment will never again be needed.

Leqembi: what the 48-month data add

In the Clarity AD longer-term follow-up, 478 participants were still receiving lecanemab at four years. Average CDR-SB worsening was 1.75 points lower than matched ADNI data and 2.17 points lower than BioFINDER data at 48 months. A dedicated analysis translated these differences into 11.0 months versus ADNI and 13.1 months versus BioFINDER, reasonably summarized as about 12 months at four years. A separate analysis reported a 34% lower relative risk of progressing to the next disease stage compared with matched ADNI data. These are longer-term comparisons with matched untreated groups, not four years of continuing placebo-controlled treatment.111213

About 12 months at four years

A dedicated analysis estimated 11.0 months versus ADNI and 13.1 months versus BioFINDER. Describing this as about 12 months is a reasonable summary. It remains an average estimate from longer follow-up, not a promise for one person.13

What the 10-Year Lecanemab Projections Mean

At CTAD 2025, Eisai and Biogen presented a model that combined longer-term Clarity AD follow-up, results from 16 monoclonal-antibody studies, and ADNI data to estimate how disease-stage timing might change over 10 years. The model estimated that continued lecanemab treatment could delay average progression:

Modeled transitionAll modeled patientsLow-amyloid subgroup at treatment start
PET below 60 Centiloids
MCI due to AD → mild AD dementia2.5 years of modeled time savings6.0 years of modeled time savings
MCI due to AD → moderate AD dementia3.5 years of modeled time savings8.3 years of modeled time savings

These are projections, not outcomes observed over 10 years. They assume that the treatment effect continues and that the comparison groups remain similar over time. The larger estimates in patients who started with less amyloid are especially uncertain and should not be presented as an expected result for one person.17

Four-year observations suggest that the average difference may continue to accumulate. Ten-year models explore what might happen if that pattern persists, but no trial has followed a randomized placebo group for 10 years.

What proportion stayed stable or remained in an earlier disease stage?

These percentages are useful, but the endpoints must be named precisely. “No worsening on CDR-SB,” “no move to the next stage,” and “remaining within early Alzheimer’s disease” are related but are not the same outcome. In particular, the published long-term reports do not provide one clean, directly comparable percentage for people who started in MCI and remained specifically in MCI with each medicine.

Patient-centered outcomeLeqembiKisunlaHow to interpret it
Remained within the broad early-AD rangeAt 4 years, about 81% remained in the MCI or mild-dementia stages, compared with about 63% in matched ADNI natural-history data.12A directly comparable absolute 3-year percentage has not been reported in the peer-reviewed long-term-extension publication.This does not mean that 81% remained specifically in MCI; some may have moved from MCI to mild dementia while still remaining within “early AD.”
Did not move to the next CDR-SB stageBy 48 months, 53.3% of lecanemab-treated participants versus 70.1% of matched ADNI participants had progressed to the next stage—equivalently, about 47% versus 30% had not.12Through 76 weeks, Kisunla reduced the relative risk of moving to the next CDR-Global stage by 37% in the overall population and by 39% in the low/medium-tau population. At 3 years, early-start treatment reduced that risk by 27% compared with delayed-start treatment; the formal reports give relative risk rather than a comparable absolute stage-retention percentage.410Relative-risk reductions describe the difference between groups; they do not reveal the exact percentage of all patients who remained in MCI.
No measured worsening on CDR-SBIn a small, selected low-tau subgroup, 69% had no worsening from baseline at 4 years; this was a post hoc exploratory analysis and does not represent all Leqembi-treated patients.1112In the prespecified low/medium-tau population, 47% had no CDR-SB worsening at 1 year, compared with 29% receiving placebo.4“No worsening on a study scale” is not identical to staying in the same clinical diagnosis or preserving every daily ability.

These percentages answer different questions. Leqembi’s four-year results describe how many participants remained within the broad early-Alzheimer’s range or avoided moving to the next stage. Kisunla’s one-year result describes how many had no worsening on a cognitive and daily-function scale, while its three-year report describes the relative risk of stage progression. They should not be treated as one interchangeable “percent who remained in MCI.”

Limits of the Long-Term Estimates

What the extension data support

  • The treatment effect did not disappear after the randomized core period.
  • The average treated and comparison curves continued to separate.
  • The pattern is consistent with a disease-modifying effect that can accumulate over time.
  • Earlier treatment appears more favorable than delayed treatment.

Why the longer-term numbers are less certain

  • The randomized placebo comparison ended after about 18 months.
  • The later analyses used matched comparison groups rather than an ongoing placebo group.
  • People who remained in long-term follow-up may differ from those who stopped.
  • “Time saved” describes an average group difference, not an individual forecast.

The first 18 months provide the strongest evidence because treatment and placebo were compared directly. The three-year Kisunla and four-year Leqembi results are also clinically useful, but they are less certain because the later comparisons were not part of a continuing randomized placebo trial.

Kisunla dosing update: The main efficacy trial used the earlier titration schedule. In a smaller safety study, the current dosing schedule lowered ARIA-E through 52 weeks from 24.9% to 16.2%. The newer schedule produced similar amyloid lowering at 24 weeks, but it was not a second placebo-controlled trial of cognitive and functional benefit.25

How the Treatments Are Given and Monitored

LeqembiKisunla
Starting treatment10 mg/kg IV every 2 weeks; about a 1-hour infusionIV every 4 weeks; about a 30-minute infusion
Dose patternSame weight-based IV dose for the first 18 months350 mg, then 700 mg, then 1,050 mg, then 1,400 mg from infusion 4 onward
Scheduled safety MRIBaseline; within about 1 week before infusions 3, 5, 7, and 14Baseline; before infusions 2, 3, 4, and 7
After the main periodContinue IV every 2 weeks, transition to IV every 4 weeks, or use weekly 360 mg under-the-skin maintenance after 18 monthsClinician may consider stopping after amyloid PET shows reduction to minimal plaque levels

Leqembi maintenance note: After 18 months, continuing every-two-week IV treatment or considering monthly IV or weekly subcutaneous maintenance are label-authorized options. The label states that transition to the less frequent maintenance regimens is supported by pharmacokinetic and pharmacodynamic modeling and that data on their long-term clinical benefit remain limited.1

Additional MRI and clinical assessment are needed whenever symptoms suggest ARIA or when a scheduled MRI shows a new abnormality. Expert recommendations may add another MRI for higher-risk patients, and treatment centers may use policies stricter than the minimum label schedule.1278


Infusion Reactions and Allergic Reactions

An infusion-related reaction is not automatically the same as a true allergy. It may involve immune activation during or after an infusion and can cause fever, chills, body aches, headache, nausea, vomiting, flushing, dizziness, blood-pressure changes, chest discomfort, reduced oxygen level, or shortness of breath.

Leqembi

26%

Infusion-related reactions occurred in 26% of treated patients in the pivotal trial. Three quarters occurred with the first infusion; most were mild or moderate, and 1% discontinued treatment because of the reaction.1

Kisunla – current titration

16%

Infusion-related reactions occurred in 16%; 88% occurred within the first four infusions, most were mild or moderate, and 2.8% discontinued. A separate broad category of hypersensitivity reactions, which included but was not limited to anaphylaxis, was reported in 8% in the 212-person modified-titration study. The original pivotal-regimen study reported 9% infusion-related reactions and 3% hypersensitivity reactions.2

Why the Kisunla rates differ: The 16% current-regimen figure and the 9% original-regimen figure came from separately conducted studies. The higher rate in Study 2 should not be interpreted as proof that the gentler titration itself caused more infusion reactions; trial populations, observation, and event ascertainment can differ across studies.2

Infusion teams may slow or stop the infusion and treat symptoms. Premedication may be used after a prior reaction. Serious hypersensitivity to either medicine is a contraindication to receiving that medicine again.12

What Exactly Is ARIA?

  • ARIA-E means edema or sulcal effusion: temporary swelling or fluid visible on MRI.
  • ARIA-H means hemosiderin deposition: new microhemorrhages or superficial siderosis, an iron-containing blood-breakdown product along the brain surface.

ARIA-E and ARIA-H can occur together, so their percentages should not be added. Some microbleeds and siderosis occur spontaneously in Alzheimer’s disease because amyloid can collect in vessel walls, a process called cerebral amyloid angiopathy. Anti-amyloid therapy raises the risk further in susceptible patients.12

Why ARIA can happen – and why MRI matters: Amyloid can also collect in the walls of small brain blood vessels. A leading explanation is that, as treatment mobilizes amyloid, these vessels may temporarily become inflamed, leaky, or more fragile, causing swelling or small areas of bleeding. A patient can feel completely normal while an MRI shows ARIA, so scheduled safety MRIs should not be skipped simply because the patient feels well.1278

ARIA-E often improves or resolves on follow-up MRI. ARIA-H is different: the active bleeding process may stabilize, but the iron deposit can remain visible. It is therefore misleading to say that every form of ARIA simply “goes away.”

Current ARIA Rates in Clinical Trials

Figures verified against current U.S. prescribing information. The rates below use the June 30, 2026 Leqembi label and the current Kisunla label. ARIA-E and ARIA-H can overlap, so their percentages must not be added.12

OutcomeLeqembi: 898 treated, 18 monthsKisunla current modified titration: 212 treated, 12 months
Any ARIA on MRI21%about 1 in 529%about 3 in 10
ARIA-E13%about 1 in 816%about 1 in 6
ARIA-H17%about 1 in 625%1 in 4
Symptomatic ARIA3% overallabout 1 in 333% symptomatic ARIA-E; less than 1% symptomatic ARIA-Habout 1 in 33 for symptomatic ARIA-E
Serious ARIA symptoms0.7%about 1 in 143No single combined percentage reported for the current regimennot reported as one combined rate
Brain hemorrhage larger than 1 cm0.7%6 of 898; about 1 in 1431% in the current-regimen studysmall study, n=212; 0.5% in the larger original-regimen pivotal trial, n=853

Study context: Leqembi’s figures come from an 18-month placebo-controlled trial; placebo rates were 9% for any ARIA, 2% for ARIA-E, 9% for ARIA-H, and 0.1% for hemorrhage larger than 1 cm. The current Kisunla figures come from a smaller active-regimen study designed primarily to compare titration schedules, not from a new placebo-controlled efficacy trial. For the rare large-hemorrhage outcome, 1% in 212 current-regimen participants is statistically less stable than 0.5% in the larger 853-person original-regimen pivotal study; both should be understood as infrequent group estimates rather than precise personal probabilities. The two drug columns should not be used to declare one medicine safer for a particular patient.125

Most ARIA Is Detected Before Symptoms Develop

Most MRI-detected ARIA causes no symptoms. In the labels, clinical symptoms resolved during observation in 79% of symptomatic Leqembi cases and in about 67% of symptomatic ARIA-E cases with the current Kisunla titration by 12 months. These figures also mean that symptoms had not resolved within the stated observation period for every patient. Serious ARIA may cause seizure, encephalopathy, focal weakness, language or vision loss, disability, or death.12

When Does ARIA Usually Occur?

Risk is front-loaded but never becomes zero. Most Leqembi ARIA-E events occurred within the first seven doses, and the label calls for enhanced vigilance during the first 14 weeks. Most Kisunla ARIA-E events occurred within the first 24 weeks. ARIA can occur later, recur, or appear after a previously normal MRI.12

How Severe Was ARIA on MRI?

With Leqembi, severe radiographic ARIA-E occurred in 1%; severe ARIA-H microhemorrhage in 3%; and severe superficial siderosis in 0.4%. With the current Kisunla titration, severe radiographic ARIA-E was not observed among 212 participants through 12 months, while severe ARIA-H microhemorrhage occurred in 2% and severe superficial siderosis in 1%. The Kisunla sample is too small for “zero observed” to be interpreted as zero true risk.12

APOE ε4: The Strongest Established Genetic Risk Factor

APOE ε4 is a common inherited variant. Two copies – one from each parent – confer the highest ARIA risk. Both labels recommend APOE testing before treatment so the risk discussion can be more individualized. Testing also reveals inherited information relevant to relatives, so the implications should be discussed before testing.12

Leqembi: Genotype-Specific Risk in the 18-Month Trial

There are two useful ways to present Leqembi risk by genotype. The current U.S. label gives a compact summary of any ARIA, symptomatic ARIA-E, and serious ARIA. The CLARITY AD subgroup analysis provides a more detailed breakdown of ARIA-E and ARIA-H. Both views are shown because they answer different questions.13

Current U.S. label summary
APOE ε4 resultParticipantsAny ARIASymptomatic ARIA-ESerious ARIA
No ε4 copies27813%About 1%About 1%
One ε4 copy47919%About 2%About 1%
Two ε4 copies14145%About 9%3%

The label uses rounded percentages. “Any ARIA” includes MRI-detected ARIA-E, ARIA-H, or both. These categories overlap and should not be added.

Detailed CLARITY AD subgroup breakdown
APOE ε4 resultARIA-E
swelling/fluid
ARIA-H
microhemorrhage or siderosis
Symptomatic ARIA-E
No ε4 copies5.4%11.9%1.4%
One ε4 copy10.9%14.0%1.7%
Two ε4 copies32.6%39.0%9.2%

These detailed rates come from the CLARITY AD genotype subgroup analysis. “Symptomatic ARIA-E” is more specific than “any symptomatic ARIA.” ARIA-E and ARIA-H may occur in the same person, so their percentages must not be summed.3

Two ε4 copies substantially increase both the chance of an MRI abnormality and the chance that brain swelling will cause symptoms. Having zero or one copy does not eliminate risk; the baseline MRI and medicines that affect bleeding also matter.

Kisunla: Why the Current Genotype Numbers Need Context

The current Kisunla dosing schedule has the most relevant safety data, but the APOE groups were small. The percentages below estimate risk during the first year, and the exact risk is less certain when only a small number of people were studied.25

Current modified titration
APOE ε4 resultParticipantsARIA-E through 52 weeksARIA-H through 52 weeksSymptomatic ARIA-E
No ε4 copies7513.8%16.4%3%
One ε4 copy11516.4%30.8%4%
Two ε4 copies2124.4%28.6%Too few to estimate reliably (0 of 21 observed)*

Only 21 people with two ε4 copies received the current dosing schedule. That group is too small to estimate the true symptomatic risk reliably.

Larger original study: a more cautious guide for two-copy risk
APOE ε4 resultParticipantsAny ARIA through 18 monthsSymptomatic ARIA-ESerious ARIA
No ε4 copies25525%4%1%
One ε4 copy45236%7%2%
Two ε4 copies14355%8%3%

This larger study used an earlier dosing schedule, so it is not the exact event rate for current dosing. It is included because the larger groups give a more reliable picture of risk by APOE status.2

How to think about risk with two APOE ε4 copies: the 0-of-21 result is too small to show the true risk. Until larger data are available, the larger original study gives a more cautious estimate: about 8% had symptomatic ARIA-E and 3% had serious ARIA. People with two copies should still be considered the highest-risk group.

Baseline MRI Findings That Raise Risk

For both medicines, the presence of at least two microhemorrhages or at least one area of superficial siderosis at baseline was associated with greater ARIA risk. The pivotal trials excluded many patients with higher-risk imaging, including more than four microhemorrhages, prior hemorrhage larger than 1 cm, vasogenic edema, and other vascular lesions; the exact exclusions differed by trial. Kisunla’s pivotal trial also excluded more than one area of superficial siderosis and severe white-matter disease.12

This matters because label rates come from a screened population. A person with more extensive cerebral amyloid angiopathy, prior hemorrhage, substantial vascular disease, or inability to undergo the required MRI monitoring may face a different or unacceptable risk. Expert recommendations and local programs may be more conservative than the label.78


Antiplatelet Medicines (Including Aspirin), Anticoagulants, and Clot-Busting Treatment

Aspirin is an antiplatelet medicine. These medication categories are not all treated the same: one antiplatelet medicine may be compatible with treatment after individual review, while chronic therapeutic anticoagulation is a greater concern.

  • A single antiplatelet medicine – most commonly low-dose aspirin – may be compatible with treatment after individual review. Most clinical-trial experience was with aspirin. Evidence is more limited for non-aspirin antiplatelet medicines and especially for two antiplatelet medicines taken together.12
  • Therapeutic anticoagulation is a greater concern. In the Leqembi pivotal trial, hemorrhage larger than 1 cm occurred in 2 of 79 patients (2.5%) receiving an anticoagulant alone or with an antiplatelet, versus none in the placebo subgroup. The group was small, so the exact risk is uncertain.1
  • Many specialty programs do not offer anti-amyloid treatment to patients who require chronic therapeutic anticoagulation, such as warfarin, apixaban, rivaroxaban, dabigatran, edoxaban, or heparin. Clot-busting treatment for possible stroke also requires special caution because ARIA can look like a stroke.78

Fatal hemorrhage after thrombolysis has occurred. The Kisunla label describes one fatal hemorrhage after a thrombolytic in the original-regimen study and one in the current modified-regimen study. In the longer-term lecanemab follow-up, two intracerebral-hemorrhage deaths occurred during lecanemab exposure: one after tPA and one in a patient receiving anticoagulation. These events are rare but real. They do not justify delaying a 911 call for possible stroke; they do make immediate disclosure of anti-amyloid treatment essential.126

Do not start or stop aspirin, another antiplatelet medicine, an anticoagulant, or any other medicine that affects clotting on your own. The treating neurologist and the clinician who prescribed the medicine should make a coordinated decision.

Suggested Medical-Alert Wording

Suggested medical-alert wording

Receiving anti-amyloid therapy: Leqembi (lecanemab) / Kisunla (donanemab). Risk of ARIA and intracerebral hemorrhage. ARIA may mimic ischemic stroke. Notify stroke/neurology team urgently before thrombolytic or antithrombotic decisions.

Symptoms That Need Prompt Evaluation

Contact the treatment team promptly for a new or worsening headache, confusion, unusual sleepiness, dizziness, imbalance, difficulty walking, visual change, nausea or vomiting, speech difficulty, weakness, numbness, seizure, or loss of awareness.

Call 911 immediately for possible serious ARIA or stroke
  • Sudden weakness or numbness, trouble speaking, loss of vision, or severe difficulty walking
  • A new seizure, loss of consciousness, or a sudden severe headache
  • Do not drive yourself and do not wait for the clinic to call back

Tell emergency staff immediately that the patient is receiving Leqembi or Kisunla. ARIA can mimic ischemic stroke and may affect thrombolytic and antithrombotic decisions. MRI is usually needed to evaluate ARIA; a routine stroke CT may not exclude it.78

What Happens If ARIA Is Found?

Management depends on symptoms and MRI severity. Some people with mild, asymptomatic ARIA may continue treatment under label-directed clinical judgment and close MRI follow-up. Symptomatic, moderate, or severe ARIA usually leads to treatment suspension; severe or recurrent events may lead to permanent discontinuation. Follow-up MRI is used to confirm that ARIA-E has resolved and ARIA-H has stabilized before a restart is considered.1278

The risk-benefit decision should be revisited after any ARIA event, any new need for anticoagulation, progression beyond the mild stage, repeated missed MRIs, or a change in the patient’s goals or treatment burden.


An Unresolved MRI Question: Brain-Volume Change

Several anti-amyloid trials have reported greater MRI-measured loss of whole-brain or hippocampal volume, or greater ventricular enlargement, in treated groups than in placebo groups. The biological meaning is unsettled. Proposed explanations range from removal of plaque, fluid, and inflammation—sometimes called amyloid-removal-related pseudoatrophy—to the possibility that some of the change reflects treatment-related tissue injury.1819

These MRI volume changes are not the same as proven accelerated neuronal loss, have not been validated as a patient-level measure of harm, and are not currently used like ARIA to decide whether a scheduled dose should be given. They nevertheless deserve transparent reporting and further study. This unresolved research question does not replace the established MRI surveillance required for ARIA.1819

What Do Real-World Data Add?

Real-world evidence is now large enough to be useful, but it answers different questions from a randomized trial. It can show whether patients complete MRIs, receive doses on schedule, remain on therapy, and experience safety events in ordinary practice. It is much less reliable for proving cognitive benefit because there is usually no randomized untreated control.

Uses of Real-World Data

  • Whether treatment is feasible outside a trial.
  • How often doses are delayed or treatment is discontinued.
  • Whether new safety patterns emerge.
  • How closely clinical populations resemble trial populations.

Limits of Real-World Data

  • That stability in one patient was caused by the medicine.
  • That lower observed ARIA rates are the true long-term rates.
  • That people who persist on treatment represent everyone who started.
  • That claims-data persistence equals effectiveness.

A 2026 U.S. claims analysis included 10,763 people with continuous healthcare encounters after at least one IV lecanemab treatment. Dosing intervals were generally close to the recommended every-two-week schedule. In a smaller subgroup of 371 people with enough follow-up to estimate longer persistence, 78.4% remained on treatment at 18 months, 71.7% at 20 months, and 67.3% at 24 months. This supports real-world feasibility and willingness to continue, but claims data do not establish cognitive effectiveness or explain why an individual stopped.15

A preliminary 2026 ALZ-NET conference abstract described 742 people receiving commercial Leqembi for an average of 8.4 months. ARIA-E was reported in 5% and ARIA-H in 7%, with symptomatic events in about 1% for each category. These findings are reassuring but should not replace the label rates in consent discussions: follow-up was shorter than the pivotal trial, the data were preliminary, and the report was a conference abstract rather than a complete peer-reviewed analysis.9

Clinical context: Real-world experience increasingly shows that organized programs can deliver treatment and monitoring at scale. For effectiveness counseling, however, the randomized trials and long-term extension analyses remain more informative than a clinician’s impression that a patient appears “stable.”

Questions to Ask Before Starting

  1. How certain are we that Alzheimer’s disease is the main cause of the symptoms and that I am still in the stage studied?
  2. What does my baseline MRI show about microbleeds, superficial siderosis, white-matter disease, or prior hemorrhage?
  3. What is my APOE ε4 result, and how does it change my personal risk?
  4. Can aspirin or another single antiplatelet be continued, and does any anticoagulant make treatment too risky?
  5. What is this program’s plan for anticoagulation, emergency clot-busting treatment, missed MRIs, recurrent ARIA, and restarting or stopping treatment?
  6. How will we judge whether the modest potential benefit remains worth the medical risk, time, cost, and burden?
  7. What should my care partner monitor, and whom should we call after hours?
CP

The role of a care partner

A care partner should know the treatment schedule, attend key visits when possible, watch for subtle confusion, gait change, headache, or unusual sleepiness, and know whom to call after hours. A care partner may notice early ARIA symptoms before the patient does.

Anti-amyloid therapy can modestly slow decline in carefully selected patients with early, amyloid-confirmed Alzheimer’s disease. The randomized trials correspond to about 5 to 6 months of average time-equivalent slowing by roughly 18 months. Longer follow-up is reasonably summarized as about seven months at three years with Kisunla and about 12 months at four years with Leqembi. These later estimates are useful but less certain because there was no continuing randomized placebo group. ARIA is common on MRI, symptoms occur in a small minority, and serious or fatal brain bleeding is rare but possible. Safe treatment requires accurate diagnosis and staging, APOE counseling, a high-quality baseline MRI, scheduled follow-up imaging, a care partner, rapid evaluation of new symptoms, and a clear plan for anticoagulation and emergency stroke care.


Related overview: This guide is the detailed benefit-risk companion to Alzheimer’s Disease Updates 2026: Blood Tests, Leqembi, Kisunla, and New Research, which covers diagnosis, blood biomarkers, symptomatic treatment, prevention research, lifestyle evidence, and the broader Alzheimer’s pipeline.

References

  1. U.S. National Library of Medicine. LEQEMBI (lecanemab-irmb) U.S. Prescribing Information. Label updated June 30, 2026; accessed July 11, 2026.
  2. U.S. National Library of Medicine. KISUNLA (donanemab-azbt) U.S. Prescribing Information. Current label accessed July 11, 2026.
  3. van Dyck CH, Swanson CJ, Aisen P, et al. Lecanemab in Early Alzheimer’s Disease. N Engl J Med. 2023;388:9-21. doi:10.1056/NEJMoa2212948. (Supplementary Appendix).
  4. Sims JR, Zimmer JA, Evans CD, et al. Donanemab in Early Symptomatic Alzheimer Disease: The TRAILBLAZER-ALZ 2 Randomized Clinical Trial. JAMA. 2023;330(6):512-527. doi:10.1001/jama.2023.13239.
  5. Wang H, Monkul Nery ES, Ardayfio P, et al. Modified titration of donanemab reduces ARIA risk and maintains amyloid reduction. Alzheimer’s & Dementia. 2025;21:e70062. doi:10.1002/alz.70062.
  6. Honig LS, Barakos J, Dhadda S, et al. Updated safety results from phase 3 lecanemab study in early Alzheimer’s disease. Alzheimers Res Ther. 2024;16:105. doi:10.1186/s13195-024-01441-8.
  7. Cummings J, Apostolova L, Rabinovici GD, et al. Lecanemab: Appropriate Use Recommendations. J Prev Alzheimers Dis. 2023;10(3):362-377. doi:10.14283/jpad.2023.30.
  8. Rabinovici GD, Selkoe DJ, Schindler SE, et al. Donanemab: Appropriate use recommendations. J Prev Alzheimers Dis. 2025;12(5):100150. doi:10.1016/j.tjpad.2025.100150.
  9. Sabbagh M, et al. Commercial Lecanemab Safety Surveillance in the ALZ-NET Registry. American Academy of Neurology Annual Meeting abstract. 2026. Preliminary conference abstract; not yet a full peer-reviewed report.
  10. Zimmer JA, Sims JR, Evans CD, et al. Donanemab in early symptomatic Alzheimer’s disease: results from the TRAILBLAZER-ALZ 2 long-term extension. J Prev Alzheimers Dis. 2026;13(2):100446. doi:10.1016/j.tjpad.2025.100446.
  11. Eisai Co., Ltd. and Biogen Inc. Early Alzheimer’s Patients Continue to Benefit from Four Years of LEQEMBI Therapy: New Clinical Data Presented at AAIC. July 30, 2025. Sponsor news release reporting the Clarity AD open-label-extension analysis.
  12. van Dyck CH, et al. Clarity AD open-label extension: efficacy and disease-stage analyses through 48 months. AAIC 2025 presentation. Sponsor-supported conference material.
  13. Dickson S, Whetten J, Hendrix S, et al. Long-Term Benefit of Lecanemab: Estimation of Time Saved Through 48 Months. CTAD 2025 poster. Sponsor-supported conference material.
  14. Dickson SP, Wessels AM, Dowsett SA, et al. “Time Saved” as a Demonstration of Clinical Meaningfulness and Illustrated Using the Donanemab TRAILBLAZER-ALZ Study Findings. J Prev Alzheimers Dis. 2023;10(3):595-599. doi:10.14283/jpad.2023.50.
  15. Eisai Co., Ltd. Results from Real-World, Long-Term Treatment Persistence with LEQEMBI in the United States Presented at AD/PD 2026. March 23, 2026. Sponsor news release describing a claims-based observational analysis.
  16. Eli Lilly and Company. KISUNLA efficacy, disease-stage, treatment-completion, and long-term-extension data. HCP website summarizing TRAILBLAZER-ALZ 2 and exploratory long-term analyses; accessed July 11, 2026.
  17. Eisai Co., Ltd. and Biogen Inc. Eisai Presents New Data on the Continued and Expanding Benefit of LEQEMBI Maintenance Treatment in Early Alzheimer’s Disease at CTAD 2025. December 4, 2025. Sponsor news release describing a 10-year modeled time-savings analysis.
  18. Belder CRS, Boche D, Nicoll JAR, et al. Brain volume change following anti-amyloid β immunotherapy for Alzheimer’s disease: amyloid-removal-related pseudo-atrophy. Lancet Neurol. 2024;23(10):1025-1034. doi:10.1016/S1474-4422(24)00335-1.
  19. Alves F, Ayton S. Missing tissue, missing data: Resolving brain volume loss caused by anti-amyloid therapies. PLoS Med. 2026;23:e1004880. doi:10.1371/journal.pmed.1004880.
  20. AlzForum. Leqembi therapeutic profile. Development history and mechanism overview; accessed July 11, 2026.
  21. AlzForum. Kisunla therapeutic profile. Development history and mechanism overview; accessed July 11, 2026.

Medical disclaimer. This article is provided for general educational and informational purposes only. It does not constitute individualized medical advice, diagnosis, treatment, a recommendation for any specific test or therapy, or informed consent. Medical information changes over time, and this article may not reflect developments occurring after its stated review date or apply to your individual circumstances. Do not disregard or delay professional medical advice, or start, stop, or change any medication or treatment, based on this article. Discuss personal medical decisions with a qualified healthcare professional who can evaluate your individual history and circumstances. Accessing or using this website does not, by itself, create a physician-patient relationship with Los Altos Neurology or any of its clinicians. Do not use this website for urgent or emergency medical concerns; call 911 or seek immediate emergency care. Although Los Altos Neurology makes reasonable efforts to provide accurate and current information, it does not guarantee that all content is complete, current, or applicable to every individual. See our full Medical Disclaimer.