The context for this work is enormous. Alzheimer’s disease accounts for roughly 60 to 80 percent of dementia cases worldwide, affecting an estimated 57 million people today, a figure projected to exceed 150 million by 2050. China alone carries nearly 17 million cases of Alzheimer’s and other dementias. For decades, treatment relied on symptomatic drugs such as acetylcholinesterase inhibitors and memantine, which ease symptoms but do nothing to halt the underlying neurodegeneration. Lecanemab changed that calculus. In the pivotal Phase III Clarity AD trial, involving 1,795 patients with early disease, the antibody slowed cognitive decline by 27 percent over 18 months and significantly reduced amyloid plaque burden on PET imaging. That success earned accelerated FDA approval in January 2023, full approval in July 2023, and Chinese regulatory approval in January 2024.

But clinical trials, however rigorous, are not real life. Trial populations are carefully selected, typically younger and healthier than the patients clinicians actually see, and rare adverse events or those with long latency periods often only surface after a drug reaches a broad population. That is precisely why the research team, led by investigators at the First Affiliated Hospital of Anhui Medical University and collaborating institutions, chose a dual approach. They prospectively followed every patient receiving lecanemab at their center from July 2024 to February 2026, and in parallel they ran a disproportionality analysis on FAERS, the FDA’s spontaneous reporting database, using two standard signal-detection metrics: the reporting odds ratio and the proportional reporting ratio.

The prospective cohort tells a relatively encouraging story. Ninety-one patients with mild-to-moderate Alzheimer’s disease were enrolled, with a median treatment duration of 185 days. Thirteen adverse events were recorded, corresponding to an exposure-adjusted incidence of 56.48 events per 100 patient-years. The most frequent were amyloid-related imaging abnormalities with hemosiderin deposition, known as ARIA-H, at 12.39 per 100 patient-years, and infusion-related reactions at 8.85 per 100 patient-years. All seven ARIA-H cases were graded mild or moderate, and none required permanent discontinuation of the drug. Notably, the cohort recorded zero cases of ARIA-E, the edema form of the abnormality, likely reflecting the team’s decision to exclude APOE ε4 homozygotes, the subgroup at highest risk, along with strict imaging-based exclusion criteria borrowed from the Clarity AD protocol.

The FAERS analysis, drawing on 2,969 reports identifying lecanemab as the primary suspect drug, largely mirrored the cohort findings. The most frequently reported events were headache, ARIA-E, ARIA-H, and infusion-related reactions, all of which appear in the drug’s official prescribing information. The signal strengths for ARIA-E and ARIA-H were extraordinarily high, with reporting odds ratios exceeding 13,000, reflecting how tightly these imaging abnormalities are associated with anti-amyloid therapy in the pharmacovigilance record. Infusion-related reactions also showed a robust signal, with a reporting odds ratio of 45.71 across 263 cases, consistent with the known immunological profile of a large-molecule exogenous protein triggering fever, chills, and blood pressure fluctuations during intravenous administration.

Then came the finding that stands out. Cerebral infarction, or ischemic stroke, is not listed in lecanemab’s prescribing information, yet the FAERS analysis revealed a strong signal with a reporting odds ratio of 22.31 across 55 reported cases. The researchers dug deeper, mapping concomitant medications and medical histories for those 55 patients. They found that 17 of them, roughly 31 percent, had at least one identifiable traditional stroke risk factor or were taking antithrombotic drugs such as aspirin or clopidogrel. Hypertension was documented in 11 patients and diabetes in two. Strikingly, 91 percent of the cases listed concomitant medications as being used for an unknown indication, underscoring how poorly baseline comorbidities are captured in spontaneous reporting systems. The authors are careful to frame this as a hypothesis-generating signal rather than proof of causation, noting that underlying cerebral amyloid angiopathy, combined with antithrombotic use and preexisting cardiovascular disease, could plausibly explain many of these ischemic events without lecanemab being the direct trigger.

Timing emerged as another key theme. A time-to-onset analysis of 898 FAERS reports with valid onset dates showed that nearly 46 percent of adverse events occurred within the first one to two months of treatment. A Weibull shape parameter below 1 classified the temporal profile as an early-failure curve, meaning the hazard of experiencing an adverse event is highest early on and declines thereafter, though a persistent fraction, about 13.7 percent, occurred more than a year after initiation. The prospective cohort showed a qualitatively similar pattern, with the highest proportion of events also clustering in the first two months. Infusion-related reactions tended to strike during the first one or two infusions, while ARIA, both the edema and the hemorrhage forms, could appear at any point in the treatment course, reinforcing the need for ongoing MRI surveillance rather than a fixed early window of vigilance.

Demographics added further nuance. In the FAERS data, patients aged 65 to 85 accounted for nearly 65 percent of reports, and in the prospective cohort the same age group made up 77 percent of those experiencing adverse events. Women were overrepresented in the spontaneous reports, consistent with the higher prevalence of Alzheimer’s disease in females. The authors suggest that this age bracket represents both the bulk of the treated population and a particularly vulnerable group, warranting prioritized monitoring of safety parameters in clinical practice.

Some findings were quietly reassuring. Despite gastrointestinal disorders ranking fifth in report volume, the disproportionality analysis yielded a protective signal, with a reporting odds ratio of 0.65, suggesting lecanemab does not meaningfully elevate the risk of nausea or diarrhea beyond what would be expected by chance. Infections and infestations, ranking seventh, similarly showed a protective signal at 0.49, indicating no significant increase in infection risk. These results diverge from what one might expect from the prescribing information and likely reflect reporting biases inherent to spontaneous systems, where only unusual or unexpected events tend to be filed.

The study is not without limitations, and the authors are candid about them. The prospective cohort is single-center and modest in size, with a median exposure of just 185 days, which limits conclusions about long-term safety beyond one year. The FAERS analysis cannot calculate true incidence rates because spontaneous reporting lacks a denominator, and the time-to-onset subset, representing about 30 percent of reports, may be biased toward medically attended events with precise dates. Still, the convergence of two independent data sources, a carefully phenotyped real-world cohort and a large pharmacovigilance database, offers a more textured picture than either alone. The overall message is one of cautious optimism: lecanemab’s safety profile in routine practice appears broadly manageable and consistent with trial data, provided clinicians remain vigilant during the first two months of therapy, monitor for ARIA throughout the entire treatment course, and interpret the emerging stroke signal with the cardiovascular context of each individual patient firmly in mind. Larger, multicenter prospective studies will be needed to confirm these signals and refine risk-mitigation strategies as anti-amyloid therapy becomes a routine part of Alzheimer’s care.

Subject of Research: Real-world safety of the Alzheimer's drug lecanemab assessed through a prospective patient cohort and FDA adverse event database signal analysis

Article Title: Safety of lecanemab in real-world patients with Alzheimer’s disease: a prospective, observational case series supported by FAERS signal analysis

Article References: Ji, H., Yang, Y., Wu, X., & Bao, E. (2026). Safety of lecanemab in real-world patients with Alzheimer’s disease: a prospective, observational case series supported by FAERS signal analysis. Frontiers in Neuroinformatics, 20, Article 1977226. https://doi.org/10.3389/fninf.2026.1977226

Image Credits: AI Generated

DOI: 10.3389/fninf.2026.1977226

Keywords: lecanemab, Alzheimer's disease, ARIA, FAERS, drug safety, pharmacovigilance, cerebral infarction, infusion-related reactions, amyloid-beta, monoclonal antibody, adverse events, real-world evidence