Safety and short-term outcomes of lecanemab for Alzheimer’s disease in China: a multicentre study

Abstract

Lecanemab is a newly approved monoclonal antibody targeting amyloid plaques for the treatment of early Alzheimer’s disease. This study aimed to evaluate the safety and short-term biomarkers and cognition changes of lecanemab in Chinese clinical practice.

This multicentre real-world study involved patients receiving lecanemab treatment across seven hospitals in China. Patients underwent comprehensive assessments before treatment. Lecanemab was administered via intravenous infusion every 2 weeks. Treatment-related symptoms were monitored through self-report, and amyloid-related imaging abnormalities were assessed via magnetic resonance imaging. Amyloid and tau biomarker changes were measured using positron emission tomography imaging or plasma testing. Follow-up cognitive assessments were evaluated after 6 months of treatment. Short-term outcomes were analysed using linear mixed-effect models, without an untreated control group.

A total of 407 patients who received at least one lecanemab infusion were involved in this study, with a mean follow-up time of 5.6 ± 3.39 months. The mean age was 68.08 years, with 67.57% of patients being female. Of the participants, 56.51% were apolipoprotein E (APOE) ε4 carriers, and 83.19% were at biological Stage C. During the study period, 22.22% of the patients experienced treatment-related symptoms, and 12.15% developed at least one amyloid-related imaging abnormality. Only four symptomatic and seven severe amyloid-related imaging abnormality cases were reported. APOE ε4 status was not related to adverse events in the Chinese population. Patients with a higher number of microhaemorrhages at baseline were more likely to develop adverse events. No significant differences in adverse events were observed between the moderate Alzheimer’s disease dementia group and the mild cognitive impairment group. By the end of the research period, 9.38% of the patients withdrew from lecanemab. After 6 months of treatment, favourable short-term outcomes in biomarkers and stable cognitive function were observed.

This study demonstrates that lecanemab treatment is feasible and well-tolerated among the Chinese population, with lower rates of adverse events and favourable short-term outcomes observed. Administration of lecanemab in the moderate Alzheimer’s disease dementia population was relatively safe and further studies are warranted.

In a multicentre real-world cohort study involving 407 patients with Alzheimer’s disease treated at seven hospitals in China, Li et al. evaluate the safety and short-term effects of lecanemab on biomarkers and cognitive function.

Introduction

Alzheimer’s disease (AD) is the predominant aetiology of dementia, currently affecting over 55 million individuals worldwide, with prevalence projected to approach 152 million cases by 2050.¹ Its defining neuropathological features are extracellular amyloid-β (Aβ) plaque deposition and intraneuronal hyperphosphorylated tau aggregation.² Lecanemab, a humanized monoclonal antibody targeting soluble Aβ protofibrils, has emerged as a significant therapeutic breakthrough.³ In the phase III Clarity AD trial, lecanemab has slowed cognitive decline by 27% in patients with early AD,⁴ leading to regulatory approval for use in mild cognitive impairment (MCI) and mild AD dementia in multiple countries.

Although lecanemab, as the first disease-modifying therapy, has demonstrated clinical benefits, its translation into real-world practice requires further consideration. In East Asian populations, where apolipoprotein E (APOE) ε4 confers a greater risk for AD and cerebral microbleeds are more prevalent,^5,6 there is concern that this group may be more prone to amyloid-related imaging abnormalities (ARIA) risk.^{7  ,8} While the Clarity AD trial in Asian subjects⁹ reported a lower incidence of adverse events (AEs), including infusion-related reactions and ARIA, than the whole group, these findings are derived from rigorously screened cohorts under controlled conditions, limiting their generalizability to routine clinical practice. The safety of lecanemab treatment in clinical practice should be further examined.

The differences in cognitive function cut-off criteria and education levels between Asian and Western populations¹⁰ also pose challenges for accurately determining the clinical stage. Patient eligibility may be assessed by simply evaluating the patients’ cognitive function to determine whether the medication can be administered. Furthermore, as the understanding and diagnosis of AD is shifting to a biologically based framework,¹¹ optimal biomarker staging in real-world settings may complicate the management strategies and efficacy prediction of lecanemab.

To address these evidence gaps, we established China’s first multicentre study of patients treated with lecanemab, with the following primary objectives: (i) to examine the safety and short-term changes of biomarkers and cognition with lecanemab in the Chinese population in a real-world clinical setting; (ii) to identify potential risk factors associated with the incidence of AEs; and (iii) to explore the use of lecanemab across a broader patient population. This investigation aims to provide an empirical foundation for the precise implementation of lecanemab across diverse healthcare systems.

Materials and methods

Study population

This multicentre study involved patients from seven hospitals in China, including Huashan Hospital Affiliated to Fudan University, First Affiliated Hospital Zhejiang University School of Medicine, Fujian Medical University Union Hospital, The First Affiliated Hospital of Wenzhou Medical University, The Second Affiliated Hospital of Soochow University, Nanjing First Hospital and Qingdao Municipal Hospital. These centres provide a nationally representative sample spanning Shanghai, Zhejiang, Jiangsu (east China), Fujian (south China) and Shandong (north China). Consecutive patients who received at least one lecanemab infusion ordered by these centres from 26 June 2024 to 26 May 2025, were included in this study. All participants underwent a comprehensive diagnostic evaluation, including medical history, clinical examination, neuropsychological assessments, routine laboratory analyses, brain MRI and biomarker testing. The diagnosis and staging of AD were based on the latest criteria.¹¹ After a thorough examination and diagnosis, patients identified as potential candidates for lecanemab referring to the appropriate use recommendations¹² and interested in the treatment were included in the unified management. The study protocol received approval from the Ethics Committee of each centre. Informed consent was obtained from all participants or their legally authorized representatives.

Treatment workflow

This is a prospectively designed cohort study with consecutive patients initiating lecanemab treatment at each centre, the inclusion workflow is shown in Fig. 1A. Candidates for lecanemab treatment were further classified into stages of AD based on their severity of cognitive impairment as rated by the Clinical Dementia Rating (CDR). All candidates had a diagnosis of symptomatic AD with a CDR of 0.5 (MCI), 1 (mild dementia) or 2 (moderate dementia). Most candidates had a Mini-Mental State Examination (MMSE) score of 20 or higher. Activities of Daily Living (ADL) scales was additionally assessed. The number of APOE ɛ4 alleles was tested in each centre using standard laboratory techniques. For patients with conditions that would have excluded them from treatment, including moderate dementia and those taking anticoagulant medications, they were informed by the doctor of the reasons why they were not suitable for lecanemab and of the serious side effects. For patients who insisted on lecanemab treatment, the neurology specialist conducted a more comprehensive evaluation and engaged in in-depth discussions with patients and their families.

Figure 1.

Workflow of the study, number of follow-up patients and biological stages of the included patients. (A) Workflow of patient inclusion and monitoring. (B) Flow of the number of patients who completed 3-month, 6-month and 9-month follow-up visits. (C) The percentages of biological stages determined by tau PET in the MCI, mild AD dementia and moderate AD dementia groups. AD = Alzheimer’s disease; ARIA = amyloid-related imaging abnormalities; CDR = Clinical Dementia Rating; CSF = cerebrospinal fluid; MCI = mild cognitive impairment; N = number of patients; PET = positron emission tomography.

Lecanemab was administered via intravenous infusion every 2 weeks at infusion centres. Patients from each centre were asked to report all symptoms that occurred during their treatment period and complete monitoring MRI, biomarker testing and cognitive assessments. Structured checklists for symptom capture were distributed to each participating centre prior to the initiation of the study. All AEs considered to be related to lecanemab treatment were recorded by self-reported symptoms or by monitoring MRI. A treatment-related symptom was defined as all symptoms related to lecanemab, and were categorized by System Organ Class (SOCs) from the Medical Dictionary of Adverse Drug Reactions and graded according to the Common Terminology Criteria for Adverse Events.¹³ An infusion-related reaction (IRR) was defined as symptoms during or after lecanemab infusion (within 24 h) due to the drugs, procedures or individual differences. The severity of symptoms was categorized as mild (does not affect daily life), moderate (affecting daily life and requiring intervention treatment) and severe (requiring strict monitoring and treatment). The patients were followed until the latest infusion or withdrawal from the treatment. For the handling of AEs, such as prophylactic treatment, and suspension and resumption of lecanemab, each centre independently made the determination based on clinical practice using the instruction guidelines for the drug.¹⁴ Patients underwent monitoring MRIs before the fifth, seventh and 14th infusions. Urgent monitoring MRI was performed on patients with sudden symptoms. For patients who agreed to regularly monitor biomarker changes, plasma was collected at baseline, 3 months and 6 months after treatment initiation, and amyloid and tau PET was conducted at baseline and 6 months after treatment initiation. Follow-up cognitive assessments were conducted 6 months after treatment initiation by the same healthcare professional. All collected follow-up biomarkers and cognitive assessment results were included in the analysis of changes after short-term treatment.

MRI

All patients underwent a pretreatment brain MRI that consisted of T1, T2, fluid-attenuated inversion recovery (FLAIR), diffusion weighted imaging (DWI) and susceptibility weighted imaging (SWI). Radiological findings including the severity of white matter disease assessed by the Fazekas score, the presence and size of infarcts, the number of microhaemorrhages and superficial siderosis were evaluated. Follow-up MRI scans were performed to assess the presence of amyloid-related imaging abnormalities with oedema/effusion (ARIA-E) and haemorrhage/haemosiderin deposition (ARIA-H) including a sagittal T1, axial DWI and T2, FLAIR and SWI. MRI sequences conducted outside centres with different scanner strengths were reviewed by a neurology specialist to determine the presence of ARIA-E and ARIA-H. Monitoring MRI scans were scheduled before the fifth, seventh and 14th infusion.

For patients who developed ARIA, the severity of ARIA was categorized using the U.S. Food and Drug Administration (FDA) guidelines.¹⁴ Concurrent ARIA-E and ARIA-H were defined as presence of ARIA-E and ARIA-H in the same monitoring period, regardless of whether they were located in the same brain region. Most people underwent a reviewing MRI 1 month after the occurrence of ARIA. Symptomatic ARIA was defined as the presence of an ARIA on regular or urgent MRI in patients presenting with clinical symptoms attributable to ARIA.

PET imaging

PET scans were performed with the Biograph mCT Flow PET/CT system (Siemens), uMI 780 PET/CT system, uMI Panorama 35S PET/CT system, uMI 790 PET/MR system (United Imaging Intelligence Co., Ltd.) and Discovery MI PET/CT system (GE HealthCare) in each centre. Either ¹⁸F-Flobetapir or ¹⁸F-Forbetaben PET scans were acquired and subjected to visual assessment for initial confirmation of Aβ deposition. Tau pathology was performed using ¹⁸F-Florzolotau, ¹⁸F-Flortaucipir or ¹⁸F-MK6240 PET images. Following acquisition, the data underwent reconstruction using a three-dimensional ordered-subset expectation maximization algorithm. The images were visually evaluated by two experienced nuclear medicine physicians. The staging of tau PET was based on previous literature¹¹ and was determined through visual read. Specifically, abnormal amyloid PET patients were categorized into Stage A (initial–no uptake on tau PET), B (early–tau PET uptake restricted to medial-temporal areas), C [intermediate–tau PET uptake in the moderate standardized uptake value ratio (SUVR) range on a neocortical region of interest (ROI)] and D (advanced–tau PET uptake in the high SUVR range in the same neocortical ROI) using tau PET results. Any disagreement between the two readers was resolved by consulting a third senior assessor. The amyloid centiloid was calculated based on a previously published method.^15,16 For tau PET SUVR in the ROIs, the cerebellar grey matter was selected as the reference region. Anatomical ROIs encompassing both cortical and subcortical regions were defined based on the Desikan–Killiany atlas.¹⁷

Biomarker testing

The CSF AD biomarker testing was conducted using standard ELISA methods (EUROIMMUN AG, CE-registered assay). Samples were collected and transported according to the requirements of the kit manufacturers and the Alzheimer’s Association’s latest guideline.¹⁸ CSF biomarkers including Aβ_1-42, Aβ_1-40, total tau and phosphorylated tau 181 were tested. All handling steps were performed according to the manufacturer’s instructions.

Blood biomarker testing was performed by single-molecule array technology. Biomarkers including Aβ₄₂, Aβ₄₀, glial fibrillary acidic protein (GFAP), neurofilament light chain (NfL) and P-Tau217 were tested. Single-molecule array technology was performed by the ALZpath P-Tau217 Advantage PLUS kit and the Human Neurology 4-Plex E (N4PE+) Advantage Plus assay kit on the HD-X platform.¹⁹

Statistical analyses

Fisher’s exact test was used for binary features and the Mann–Whitney U-test was used for quantitative features to analyse the clinical features, grouped by cognitive stages. Continuous variables were summarized using means and standard deviations (SDs) and categorical data were expressed as counts and percentages. Univariate and multivariable logistic regression analyses were performed to identify factors associated with the occurrence of AEs. Variables with a P-value <0.1 in univariate logistic analysis, as determined by the likelihood ratio test, were entered into the final multivariable model. Sensitivity analysis was performed after excluding those with superficial siderosis and medium-sized infarcts at baseline. To estimate changes in amyloid centiloid, tau SUVR and plasma biomarkers, linear mixed-effects models with random slopes and intercepts were adjusted for sex and age at first infusion. For cognitive assessments, the models were additionally adjusted for baseline severity of cognitive impairment, education, study centre, and concomitant use of acetylcholinesterase inhibitors and memantine. False discovery rate (FDR) correction was applied for multiple testing. Statistical analyses were performed using R, version 4.4.1, with all tests two-sided at a 5% level of significance.

Results

Demographics and clinical characteristics of patients

A total of 407 AD patients initiating lecanemab infusion were included in the overall study sample. The mean follow-up time was 5.6 ± 3.39 months (excluding withdrawals and dropouts). The number of patients who completed 3-month, 6-month and 9-month follow-up visits, as well as dropouts, is shown in Fig. 1B. The number of patients diagnosed with mild cognitive impairment, mild dementia and moderate dementia was 239, 136 and 32, respectively (Table 1). The mean age was 68.08 years, with 67.57% of patients being female. There were no significant differences in age and gender among the three groups. The mean education period was 9.57 ± 4.71 years, with the moderate dementia group having the lowest years of education (7.00 ± 4.56 years). The average course from symptom onset was 2.65 ± 1.54 years), with the moderate dementia group sharing the longest disease course (4.12 ± 1.88 years). A total of 56.51% of the patients were APOE ɛ4 carriers (43.98% heterozygous, 12.53% homozygous). Comorbidities of hypertension, diabetes and dyslipidaemia were observed in 32.68%, 13.51% and 28.26% of the patients. Most patients were using symptomatic medications in combination with lecanemab (70.52% acetylcholinesterase inhibitor, 23.83% N-methyl-D-aspartic acid receptor antagonist). Three patients (0.74%) were receiving anticoagulant medications, and 35 patients (8.60%) were taking antiplatelet agents. Characteristics of each centre are demonstrated in Supplementary Table 1.

Table 1.

Baseline characteristics of patients

Characteristic	Total	MCI due to AD	Mild AD dementia	Moderate AD dementia	P
Number of patients	407	239	136	32	–
Age at first infusion (years), mean ± SD	68.08 ± 8.71	68.08 ± 8.65	68.38 ± 8.39	66.84 ± 10.57	0.669
Female, n (%)	275 (67.57)	159 (66.53)	92 (67.65)	24 (75.00)	0.630
Education (years), mean ± SD	9.57 ± 4.71	10.49 ± 4.46	8.57 ± 4.79	7.00 ± 4.56	<0.001
Time since onset of symptoms (years), mean ± SD	2.65 ± 1.54	2.38 ± 1.42	2.76 ± 1.45	4.12 ± 1.88	<0.001
ApoE ɛ4 status, n (%)	–	–	–	–	0.024
ɛ4 Non-carrier	156 (38.33)	89 (37.24)	48 (35.29)	19 (59.38)	–
ɛ4 Heterozygotes	179 (43.98)	110 (46.03)	62 (45.59)	7 (21.88)	–
ɛ4 Homozygotes	51 (12.53)	32 (13.39)	17 (12.50)	2 (6.25)	–
Unknown	21 (5.16)	8 (3.35)	9 (6.62)	4 (12.50)	–
MMSE score, mean ± SD	20.20 ± 5.09	22.73 ± 3.60	17.52 ± 4.20	12.17 ± 4.40	<0.001
Concomitant medications, n (%)	–	–	–	–	–
Acetylcholinesterase inhibitor	287 (70.52)	168 (70.29)	92 (67.65)	27 (84.38)	0.174
Memantine	97 (23.83)	33 (13.81)	45 (33.09)	19 (59.38)	<0.001
Antiplatelet	35 (8.60)	17 (7.11)	16 (11.76)	2 (6.25)	0.268
Anticoagulant	3 (0.74)	1 (0.42)	2 (1.47)	0 (0.00)	0.456
Baseline Fazekas score, n (%)	–	–	–	–	0.018
0	27 (6.63)	14 (5.86)	10 (7.35)	3 (9.38)	–
1	238 (58.48)	150 (62.76)	72 (52.94)	16 (50.00)	–
2	89 (21.87)	52 (21.76)	32 (23.53)	5 (15.62)	–
3	19 (4.67)	6 (2.51)	12 (8.82)	1 (3.12)	–
Unknown	34 (8.35)	17 (7.11)	10 (7.35)	7 (21.88)	–
Baseline infarcts, n (%)	–	–	–	–	<0.001
≥1 Small infarcts (<10 mm)	75 (18.43)	49 (20.50)	23 (16.91)	3 (9.38)	–
1 Medium-sized infarct (10–20 mm)	6 (1.47)	2 (0.84)	1 (0.74)	3 (9.38)	–
Baseline microhaemorrhage number, n (%)	–	–	–	–	0.040
0	255 (62.65)	145 (60.67)	91 (66.91)	19 (59.38)	–
1	66 (16.22)	45 (18.83)	19 (13.97)	2 (6.25)	–
2–4	58 (14.25)	34 (14.23)	19 (13.97)	5 (15.62)	–
5–9	6 (1.47)	4 (1.67)	2 (1.47)	0 (0.00)	–
Unknown	22 (5.41)	11 (4.60)	5 (3.68)	6 (18.75)	–
Baseline superficial siderosis, n (%)	3 (0.74)	2 (0.84)	1 (0.74)	0 (0.00)	0.014

Fisher exact test was used for binary features and Mann–Whitney U-test was used for quantitative features to analyse the clinical features. Continuous variables were summarized using means and SDs and categorical data were expressed as counts and percentages. AD = Alzheimer’s disease; APOE = apolipoprotein E; MCI = mild cognitive impairment; MMSE = Mini-Mental State Examination; SD = standard deviation.

Most patients underwent amyloid PET [378 (92.87%)] for confirmation of amyloid pathology, while 23 patients (5.65%) underwent CSF AD biomarker testing. Four patients (0.98%) underwent both PET and CSF examinations. Two patients (0.49%) only underwent a high-accuracy blood AD biomarker test for amyloid pathology. Notably, among 382 patients who underwent PET imaging, 243 patients also underwent simultaneous tau PET for pathological staging of AD. Most patients in the MCI (85.1%) and mild dementia (84.8%) groups were at Stage C, while 45.5% of the moderate dementia patients receiving lecanemab treatment were at Stage C (Fig. 1C).

Treatment-related symptoms

Of the 407 patients, two were lost to follow-up regarding their most recent infusion. In the remaining 405 patients with available infusion data, 90 (22.22%) experienced treatment-related symptoms during the study period. Numbers and AEs in participants who completed 3-month, 6-month and 9-month follow-up visits are shown in Supplementary Table 2. The actual number of patients finishing each infusion, with the percentage and severity of treatment-related symptoms, is shown in Fig. 2A. Most treatment-related symptoms occurred after the first infusion, and the incidence rate sharply decreased after the second infusion. Symptoms were usually rated as mild or moderate, with only three severe cases reported. Except for general disorders normally caused by IRRs, the most affected system involved nervous system disorders, followed by gastrointestinal disorders. Commonly reported symptoms included pyrexia, dizziness, chills, vomiting and headache (Fig. 2B). Rarely reported symptoms of abnormal liver function indicators and sudden deafness were observed. Details of the rare symptoms are shown in the Supplementary material. The three serious adverse reactions were sudden deafness, lower limb thrombosis and severe symptomatic ARIA.

Figure 2.

Characteristics of treatment-related symptoms over the treatment period. (A) The percentage of patients with treatment-related symptoms in 405 patients who were not lost to follow-up. The upper number indicates the exact number of patients finishing each infusion. (B) The percentage of symptoms reported by patients with treatment-related symptoms is shown, with bar colours representing different organ systems categorized by systemic organ categories. (C) The Venn diagram shows the situations in which patients experienced their first IRR. The blue circle indicates patients who received medication treatment for IRRs. The pink circle represents patients who experienced IRR during the first infusion. The yellow circle indicates all patients who had experienced IRRs. The green circle represents IRRs that were alleviated within 24 h. (D) Situations of IRRs in 37 patients who received at least one infusion after their first IRR. IRR = infusion-related reaction; n = number of patients.

IRRs were reported in 41 patients (10.12%), with the most common symptoms being pyrexia and chills, and approximately half of these reactions (21/41) were mild. Most patients experienced IRRs in the first infusion, except for two patients whose infusion-related blood pressure changes occurred in subsequent infusions. Drug intervention was carried out in 23 patients with IRRs; 27 (65.85%) patients recovered within 24 h (Fig. 2C). Of 37 patients who underwent at least one infusion after the first IRR, nine patients (24.32%) experienced more than one instance of IRR, and eight of them experienced moderate IRRs during the first infusion. Among the 14 patients whose first IRRs lasted more than 24 h, 50% experienced similar IRRs during subsequent treatments. In contrast, the recurrence rate was 7.4% among patients whose first IRR resolved within 24 h (Fig. 2D).

At the end of the study period, seven patients arranged prophylactic treatment after experiencing IRRs. The characteristics, symptoms and prophylactic treatments are listed in Supplementary Table 3. In general, an oral antihistamine drug or glucocorticoid therapy was administered by the neurologists based on the symptoms and severity of IRRs. Of note, five patients with more complex and persistent symptoms still presented milder but similar IRR symptoms on the subsequent infusion.

Baseline MRI and ARIA findings

Before starting lecanemab infusion, patients underwent MRI scans to evaluate their cerebrovascular risk. On baseline brain MRI, 81 (19.90%) patients had lacunar infarctions, including 6 (1.47%) with medium-sized infarcts. The majority of patients (85.02%) had evidence of white matter disease, with a Fazekas score of either 1 [238 (58.48%)], 2 [89 (21.87%)] or 3 [19 (4.67%)]. A total of 130 patients (31.94%) reported baseline microhaemorrhage and three (0.74%) patients had superficial siderosis (Table 1).

A total of 321 patients who reached the point of monitoring scans with baseline MRI and monitoring MRI available were included in the study assessing ARIA risk; 307, 250 and 146 patients completed monitoring MRI before the fifth, seventh and 14th infusion, respectively. Among them, 39 patients (12.15%) developed at least one ARIA finding. Sixteen had ARIA-E with or without ARIA-H and 23 had isolated ARIA-H (Table 2 and Fig. 3A). The incidence of ARIA was 7.17%, 10.99% and 10.13% in patients who completed 3-month, 6-month and 9-month follow-up visits (Supplementary Table 2). Isolated ARIA-H increased after 6 months (3.6% after 2 months and 4.8% after 6 months) (Fig. 3B). Most ARIA were asymptomatic [35/39 (89.74%)] and were radiographically mild [23/39 (58.97%)]. Concurrent ARIA-E and ARIA-H were present in 12 patients, with seven moderate and severe cases demonstrating concurrent ARIA-E and ARIA-H in the same brain region (Fig. 3C). For four symptomatic ARIA cases, ARIA was detected in three patients after urgent MRIs, and one patient was diagnosed on routine monitoring MRI and reported aggravated symptoms after being contacted about the MRI results. Two mild symptomatic cases were dizziness with mild ARIA-H. A total of 11 urgent MRIs were ordered for patients with symptoms that may be associated with ARIA and three detected ARIA.

Table 2.

Occurrence of ARIA in different cognitive groups

	Total	MCI due to AD	Mild AD dementia	Moderate AD dementia	P
n	321	191	112	18	–
Any ARIA, n (%)	39 (12.15)	23 (12.04)	14 (12.50)	2 (11.11)	0.984
Symptomatic ARIA, n (%)	4 (1.25)	1 (0.52)	3 (2.68)	0 (0.00)	0.234
ARIA-E, n (%)	16 (4.98)	10 (5.24)	5 (4.46)	1 (5.56)	0.950
Mild	10 (3.12)	7 (3.66)	3 (2.68)	0 (0.00)	–
Moderate	4 (1.25)	2 (1.05)	1 (0.89)	1 (5.56)	–
Severe	2 (0.62)	1 (0.52)	1 (0.89)	0 (0.00)	–
Isolated ARIA-E, n (%)	4 (1.25)	4 (2.09)	0 (0.00)	0 (0.00)	0.252
Mild	2 (0.62)	2 (1.05)	0 (0.00)	0 (0.00)	–
Moderate	1 (0.31)	1 (0.52)	0 (0.00)	0 (0.00)	–
Severe	1 (0.31)	1 (0.52)	0 (0.00)	0 (0.00)	–
ARIA-H, n (%)	35 (10.90)	19 (9.95)	14 (12.50)	2 (11.11)	0.789
Mild	23 (7.17)	15 (7.85)	6 (5.36)	2 (11.11)	–
Moderate	7 (2.18)	2 (1.05)	5 (4.46)	0 (0.00)	–
Severe	5 (1.56)	2 (1.05)	3 (2.68)	0 (0.00)	–
Isolated ARIA-H, n (%)	23 (7.17)	13 (6.81)	9 (8.04)	1 (5.56)	0.889
Mild	16 (4.98)	10 (5.24)	5 (4.46)	1 (5.56)	–
Moderate	4 (1.25)	2 (1.05)	2 (1.79)	0 (0.00)	–
Severe	3 (0.93)	1 (0.52)	2 (1.79)	0 (0.00)	–
Concurrent ARIA-E and ARIA-H, n (%)	12 (3.74)	6 (3.14)	5 (4.46)	1 (5.56)	0.772
Mild	5 (1.56)	5 (2.62)	0 (0.00)	0 (0.00)	–
Moderate	4 (1.25)	0 (0.00)	3 (2.68)	1 (5.56)	–
Severe	3 (0.93)	1 (0.52)	2 (1.79)	0 (0.00)	–

Fisher exact test was used to analyse the clinical features. AD = Alzheimer’s disease; ARIA = amyloid-related imaging abnormalities; ARIA-E = ARIA with oedema/effusion; ARIA-H = ARIA with haemorrhage/haemosiderin deposition; MCI = mild cognitive impairment.

Figure 3.

Characteristics of ARIA over the treatment period. The analyses of ARIA included 321 patients who received baseline MRI and at least one monitoring MRI scan. (A) The proportion of patients with any ARIA-E, isolated ARIA-H or any ARIA is shown as a function of the time since first infusion. The numbers below indicate the population at risk at the current time point, and the numbers in parentheses represent the number of patients who had experienced the event. Vertical ticks on the curve represent patients censored at the longest follow-up time. (B) The percentage of ARIA events reported in the monitoring MRIs before the fifth, seventh and 14th infusion. (C) The number of cases is shown according to the maximum radiographic severity of ARIA-H and ARIA-E. The size of the circle corresponds to the number of patients. ARIA = amyloid-related imaging abnormalities; ARIA-E = ARIA with oedema/effusion; ARIA-H = ARIA with haemorrhage/haemosiderin deposition.

Factors associated with adverse events

We further explored factors that may be associated with the risk of AEs. By grouping patients by different APOE ɛ4 status and the severity of cognitive impairment, we observed that there was no higher risk of all AEs in moderate AD dementia patients, while the incidence of IRRs was higher (moderate AD dementia, 21.88%; mild AD dementia, 5.93%; MCI due to AD, 10.92%) (Table 3). After univariable and multivariable linear regression models, we demonstrated that over one microhaemorrhage at baseline was associated with higher rates of AEs [OR (95% CI): 2.55 (1.31–4.95), P = 0.006] and IRRs [OR (95% CI): 2.23 (1.03–4.68), P = 0.036]. Usage of an antiplatelet [OR (95% CI): 2.82 (1.02–7.55), P = 0.039] was independently associated with risk of AEs (Supplementary Tables 4 and 5). Sensitivity analysis demonstrated similar results.

Table 3.

Occurrence of adverse events and change in administration in different cognitive groups

	Total	MCI due to AD	Mild AD dementia	Moderate AD dementia	P
n	405	238	135	32	–
Treatment-related AE, n (%)	113 (27.90)	72 (30.25)	30 (22.22)	11 (34.38)	0.175
Treatment-related AE (other than asymptomatic ARIA), n (%)	90 (22.22)	59 (24.79)	22 (16.30)	9 (28.12)	0.117
Infusion-related reactions, n (%)	41 (10.12)	26 (10.92)	8 (5.93)	7 (21.88)	0.022
Treatment adjustment, n (%)	–	–	–	–	–
Withdrawal from treatment	38 (9.38)	15 (6.30)	17 (12.59)	6 (18.75)	0.022
Treatment interruption	29 (7.16)	20 (8.40)	7 (5.19)	2 (6.25)	0.500
AEs leading to treatment adjustment, n (%)	–	–	–	–	–
Leading to treatment withdrawal	21 (5.19)	10 (4.20)	9 (6.67)	2 (6.25)	0.564
Leading to treatment interruption	20 (4.94)	13 (5.46)	5 (3.70)	2 (6.25)	0.707

Fisher exact test was used to analyse the clinical features. AD = Alzheimer’s disease; AE = adverse events; ARIA = amyloid-related imaging abnormalities; MCI = mild cognitive impairment.

After grouping by different APOE ɛ4 status and the severity of cognitive impairment, there were no significant differences in the risk of ARIA among these groups (Table 2 and Supplementary Table 6). In univariable and multivariable linear regression analysis, compared with those without microhaemorrhage, occurrence of microhaemorrhages at baseline was associated with the risk of ARIA [baseline one microhaemorrhage: odds ratio (OR), 95% confidence interval (CI): 4.96 (2.06–11.97), P < 0.001; baseline over one microhaemorrhage: OR (95% CI): 4.98 (2.10–11.83), P < 0.001]. For isolated ARIA-H, baseline superficial siderosis is detrimental [OR (95% CI): 15.68 (1.17–385.74), P = 0.040], while the occurrence decreased with longer disease course [OR (95% CI): 0.63 (0.39–0.92), P = 0.033]. However, a higher number of baseline microhaemorrhages was not found to correlate with an increased risk of isolated ARIA-H (Supplementary Tables 7–9).

Change in lecanemab usage

During this study period, 38 patients (9.38%) withdrew from lecanemab treatment, 29 patients (7.16%) experienced short-term discontinuation and two patients were lost to follow-up. Patients withdrew from lecanemab treatment for the following reasons: 10 for asymptomatic and two for symptomatic ARIA; nine for treatment-related symptoms; 15 for subjective reasons including concerns about side effects, dissatisfaction with the therapeutic effect (worsened cognition), economic reasons and inconvenient transportation; and two for unrelated health issues. Of all withdrawn patients, 71.05% were female. Patients who experienced ARIA had a higher withdrawal rate (Fig. 4A). The moderate AD dementia group had a higher withdrawal rate (Fig. 4B), but most of them withdrew for subjective reasons, including worsened cognition leading to poor medication adherence (Table 3). No statistical difference was found in the percentage of withdrawal and treatment interruption due to AEs in different cognitive groups. Characteristics and symptoms of patients who withdrew due to AEs are listed in Supplementary Table 10. Four patients withdrew after the first infusion due to intolerance of IRRs and two patients withdrew for aggravation of AD-related symptoms. Of 29 patients with short-term discontinuation, most of them (14/29) involved temporary discontinuation of medication due to mild or moderate ARIA, and nine patients experienced irregular administrations for subjective reasons or unrelated health issues.

Figure 4.

Time-to-withdrawal curves of the patients. (A) Time-to-withdrawal curves in all patients with follow-up (N = 405), in patients who experienced treatment-related symptoms (n = 90) and in patients who developed ARIA (n = 39). (B) Time-to-withdrawal curves in MCI (n = 238), mild AD dementia (n = 135) and moderate AD dementia (n = 32) groups. AD = Alzheimer’s disease; ARIA = amyloid-related imaging abnormalities; MCI = mild cognitive impairment.

Among 39 patients who experienced ARIA, after the first findings of ARIA, 16 patients continued treatment, 14 patients suspended administration after the determination of ARIA, six patients withdrew for moderate or severe ARIA and three patients withdrew for concerns about the progression of ARIA. In patients who continued treatment, 14 patients observed stable or mild progressive ARIA in the following monitoring MRI, and the other two patients experienced newly discovered ARIA-E or severe ARIA-H. Of patients who withdrew or suspended treatment, monthly monitoring MRI demonstrated stable ARIA-H or relieved ARIA-E after interruption of infusion. In 10 patients who had already restarted treatment, only one patient was found to have experienced a worsening ARIA and withdrew from the treatment. Representative MRI images of the patients who experienced severe ARIA-E and restarted treatment after ARIA relief are shown in Supplementary Fig. 1.

Biomarker and cognition monitoring

We further explored the short-term treatment outcomes of lecanemab by evaluating changes in amyloid biomarkers and cognitive function in patients who received lecanemab infusion for over 6 months. Given the limited number of patients completing these tests, we first compared baseline characteristics and AEs between patients with and without tau PET and plasma biomarker assessments to evaluate generalizability (Supplementary Tables 11 and 12).

In 39 patients who had completed baseline and 6 months amyloid PET CT, we observed a declining trend in the amyloid centiloid from baseline to 6 months [β (95% CI): −26.60 (−32.19, −21.02), P < 0.001] using a linear mixed-effects model, adjusted for sex and age at the first infusion (Fig. 5A). Since some patients’ amyloid PET were performed by PET MR, to carry out a thorough assessment of amyloid changes, the rate of amyloid SUVR changes were evaluated. The linear regression model demonstrated a 12.47% decrease in amyloid PET SUVR after 6 months of treatment [95% CI: (−14.85, −10.08), P < 0.001; Supplementary Fig. 2]. Supplementary Fig. 3 demonstrated a case whose amyloid deposit detected by PET was close to negative after 6 months of treatment. Among 26 patients with contemporaneous follow-up tau PET, an increased tau SUVR in the cerebral cortex including inferior temporal, entorhinal and parahippocampal regions was demonstrated, and SUVR in the anterior cingulate regions decreased (Fig. 5B and Supplementary Table 13).

Figure 5.

Biomarkers and cognition changes between baseline and 6 months. (A) Amyloid centiloid value at baseline and 6 months after treatment. A linear mixed-effects model, adjusted for age at first infusion and sex, was applied to evaluate the differences. (B) The cortical and subcortical t-values of follow-up tau SUVR compared with baseline using a linear mixed-effects model, adjusted for age at first infusion and sex. (C) The results of plasma biomarker changes by a linear mixed-effects model, adjusted for age at first infusion and sex. (D) Paired box plots showing changes in MMSE between baseline and 6 months. (E) Paired box plots of ADL changes. (F) Alluvial plot illustrates categorical transitions in CDR-G between baseline and 6 months. ns = not significant; **P < 0.01; ***P < 0.001. Aβ = amyloid-β; ADL = Activities of Daily Living; CDR-G = Clinical Dementia Rating Global Score; GFAP = glial fibrillary acidic protein; MMSE = Mini-Mental State Examination; N = number of patients; NfL = neurofilament light chain; SUVR = standardized uptake value ratio.

The decrease of plasma P-Tau217 was observed after 3 months of lecanemab treatment [β (95% CI): −0.16 (−0.27, −0.054), P = 0.006], and the P-Tau217 level met a more significant decline after 6 months [β (95% CI): −0.26 (−0.36, −0.17), P < 0.001]. The P-Tau217/Aβ₄₂ levels also decreased after 3 months [β (95% CI): −0.23 (−0.36, −0.10), P = 0.001] and 6 months [β (95% CI): −0.29 (−0.40, −0.17), P < 0.001]. Moreover, plasma GFAP levels demonstrated a significant decrease after 3 months [β (95% CI): −24.55 (−37.87, −11.17), P = 0.001] and remained stable after 6 months (β (95% CI): −22.31 (−33.82, −10.84), P < 0.001]. There was no significant difference in plasma NfL and Aβ_42/40 levels after 3 months or 6 months (Fig. 5C and Supplementary Table 14).

In 182 patients who completed 6 months of treatment, about half of the patients had follow-up cognitive assessments [MMSE: 61.53% (112/182), CDR: 39.56% (72/182), ADL: 25.27% (46/182)]. Changes in MMSE, CDR and ADL are shown in Fig. 5D–F. To initially determine the changes in cognitive performance after 6 months of lecanemab treatment, we compared the baseline and follow-up scores of MMSE, CDR and ADL using linear mixed-effects models. After adjusting, a stable cognitive function was demonstrated with no significant changes in MMSE, CDR and ADL scale after FDR correction (Supplementary Table 15).

Discussion

In this study, we established a real-world setting to evaluate the safety and short-term changes of biomarkers and cognition with lecanemab in the Chinese population. Chinese AD patients adhered to lecanemab treatment with a withdrawal rate of 9.38%. Patients selected for lecanemab administration based on rigorous criteria achieved favourable benefits, with a relatively low rate of serious AEs. Throughout the study period, 22.22% of the patients experienced treatment-related symptoms, and 12.15% developed at least one ARIA. APOE ɛ4 status was not related to AEs in the Chinese population, while a higher total number of microhaemorrhages at baseline was associated with an increased risk of AEs, pointing to the need for closer monitoring. Except for higher rates of IRRs in the moderate dementia group compared with the MCI group, no significant differences in AEs were observed. However, these patients may discontinue treatment due to poor medication adherence resulting from worsening cognition. After 6 months of treatment, the patients showed favourable amyloid pathology changes with stable cognitive function, while tau pathology continued to progress.

Compared with treated participants from the whole group and the Asian region of the Clarity AD trial,^4,9 our patients receiving lecanemab were slightly younger and more cognitively impaired, with lower MMSE scores. Age is a major determinant of AD prevalence in China, with a peak incidence in individuals aged 75 years and above.²⁰ A major reason for the involvement of a younger group of patients was the payment method in China. Unlike the payment methods in Western countries, where most payments may be covered by medical insurance,²¹ Chinese patients have to pay for lecanemab entirely out of their own pockets. Young patients have a more urgent need to maintain good cognitive function for a longer period, while older patients tend to allocate their funds for life-improving methods such as nursing homes. The huge economic burden also restricts the use of lecanemab in economically disadvantaged areas. Moreover, adhering to a 2-week hospital visit schedule is a challenge for patients with poor transportation. However, the advantage in China is that patients can quickly go from diagnosis to administration. They are usually able to complete the diagnostic testing within 2 weeks of the initial visit, and start treatment within 1 week after the diagnosis.

Although the study duration of 11 months was relatively short, the occurrence rates of treatment-related AEs (27.9% versus 29.5%), ARIA-E (5.0% versus 6.2%) and ARIA-H (10.9% versus 14.4%) were consistent with the Clarity AD Asian Region trial. In contrast to real-world studies in Western populations,^22-24 the overall incidence of AEs in the Chinese cohort was lower, both for treatment-related symptoms and ARIA. However, a higher proportion of patients discontinued treatment for subjective reasons, leading to a withdrawal rate similar to that seen in Western populations. No death and only one macrohaemorrhage were reported during the study period. Overall, lecanemab demonstrated good tolerability in Chinese AD patients.

In terms of symptoms and the percentage of treatment-related AEs, this study showed consistency with previous findings on anti-Aβ monoclonal antibodies.²⁵ Notably, rare symptoms, including abnormal liver function indicators and sudden deafness, were observed. However, since these symptoms were discovered between the two administration periods rather than during the administration process, we cannot determine the correlation between these symptoms and lecanemab, and the possibility of complications cannot be ruled out. IRRs were the most frequently reported AEs during the initial infusion and were predominantly mild. Mild IRRs that resolved within 24 h rarely recurred with subsequent infusions. In contrast, patients who experienced moderate IRRs lasting over 24 h may require prophylactic treatment before lecanemab infusion. Choices of prophylactic treatment include antihistamine, acetaminophen or glucocorticoid therapy.²⁶ Preventive measures were generally effective, though severe conditions requiring glucocorticoid therapy were more likely to recur despite prophylaxis.

The majority of ARIA findings were asymptomatic and usually present as isolated ARIA-E or ARIA-H. ARIA-H findings were also noted in some patients in the placebo arm in clinical trials.^4,27 This suggests that the neuroimaging manifestations observed may not be solely attributable to treatment-related effects, but could also reflect inherent disease characteristics or distinctive neurovascular responses. Findings on associations between baseline microhaemorrhages and incidence of isolated ARIA-H support this idea. Interestingly, our study found that APOE ɛ4 carriers did not exhibit a higher risk of ARIA, despite the genotype being recognized as a major risk factor for ARIA.^7,28 This may indicate that the response to anti-amyloid treatment varies across ethnic groups and underlying gene mechanisms, with the Chinese population displaying a distinct susceptibility profile. As APOE ɛ4 carriers may have the same or better response to amyloid-targeting therapies,²⁹ with better tolerance, the Chinese population may benefit more from monoclonal treatment. In patients with a higher number of microhaemorrhages at baseline, there was a strong association with the occurrence of AEs, and therefore vigilant MRI monitoring is encouraged for greater clinical benefit. Most patients who developed ARIA did not withdraw from treatment. Monthly monitoring MRI revealed that most ARIA-H were stable, and ARIA-E can be resolved. Concerned about the progression of ARIA, some patients temporarily suspended administration and resumed once ARIA stabilized or resolved. Despite this, these patients showed no further progression of ARIA on subsequent MRI. Whether the suspension may affect the efficacy needs to be further studied.

We are aware that there are instances where lecanemab is initiated beyond the recommended use, such as in patients with moderate dementia, anticoagulant use and widespread cerebrovascular disease. These patients nonetheless initiated therapy, reflecting strong willingness to pursue treatment despite acknowledging potential AEs and uncertain efficacy. Notably, 45.5% of the moderate dementia patients were at Stage C of the tau PET biological stage, indicating a mismatch between clinical and pathological staging; possibly caused by less neurodegeneration,³⁰ thus suggesting the potential benefit of lecanemab treatment. Although previous Aβ monoclonal antibody trials failed to show clinical benefit in mild to moderate AD,³¹ these drugs were terminated for failing to achieve the expected therapeutic effects. Lecanemab has not been studied for moderate dementia. The inclusion of patients outside the scope of the clinical trial gives strength to the dataset as it provides evidence of the safety and efficacy profile in a more lenient real-world sample rather than simply recapitulating the trial inclusion criteria. This is particularly important to clinicians counselling patients about treatment who may be on the cusp of moderate AD or who are taking anticoagulants. In our study, safety outcomes were comparable to those in MCI, apart from a higher rate of IRRs. While withdrawal rates due to AEs did not differ across different cognitive groups, the moderate AD dementia group was observed to have a higher withdrawal rate due to worsened cognition and intolerance of the regular administration. Given the small sample size and limited follow-up, our data do not provide sufficient evidence to evaluate efficacy in moderate AD dementia. It remains inappropriate to make a conclusion on whether these patients can be widely treated with anti-amyloid monoclonal antibodies and the external validity is relatively low.

In preliminary analyses on changes in biomarkers, we examined the changes in PET and plasma biomarkers after 6 months of treatment. As expected, a reduction in brain amyloid burden was observed at 6 months by PET imaging. Consistent with Clarity AD and TRAILBLAZER-ALZ,³² a clinical trial of another marketed monoclonal antibody drug Donanemab, a decrease in plasma P-Tau217 and GFAP levels can be observed 3 months after lecanemab treatment and continued at 6 months. No significant change was observed for NfL levels. As clearance of NfL in humans is very slow, failure to observe changes in NfL was not a surprise. Changes in Aβ42/40 demonstrated a significant increase in the extended analysis of Clarity AD data, but were not replicated in our study. A longer-term follow-up is needed to clearly establish the changes. Cortical tau SUVR, however, showed an increase, and current evidence remains limited regarding whether lecanemab slows tau accumulation compared with non-treated patients. No public data on the effect of lecanemab treatment on tau-PET outcomes is yet available. The tau results here suggested that lecanemab treatment may not reverse or halt tau progression, and the effect on P-Tau markers may not be extrapolated to be an effect on tau tangle pathology. With TRAILBLAZER-ALZ phase 2 results demonstrating a decreased tau load,³³ the effect of amyloid plaque-targeting monoclonal antibodies on tau pathology is still pending.

Though over half of the patients had improved scores in MMSE after 6 months of treatment, there was no statistical difference in all cognitive assessments. The improving trend differed from the Clarity AD and real-world study results,²² which demonstrated a declining trend in cognitive assessment score. Practice effect on repeated assessments in this open-label setting may be one reason.³⁴ Moreover, some of the patients included were newly diagnosed with AD, and the initial use of symptomatic improvement drugs contributed to the improvement in cognitive function. Restricted by the limited sample size of follow-up biomarker testing and cognitive assessments, the correlation between clinical and biological improvement and factors associated with better biological improvement was not detected in this study. Further study on this population is expected to explore which subgroups will benefit the most from the application of lecanemab treatment.

This study represents the first real-world analysis of lecanemab usage across multiple regions of China, comprising a relatively large sample size and comprehensive documentation of complications. The data and results are expected to be used for policymaking, service planning and meta-analyses in the future. However, several limitations must be acknowledged. First, the relatively short follow-up period and small sample size may lead to an underestimation of AEs and hinder the evaluation of changes in biomarkers and cognition. These factors also made it difficult to draw a solid conclusion, including the relation between APOE ɛ4 carriers and the risk of ARIA. Still, with ongoing follow-up, we expect updated data on a larger study group at the 12-month and 18-month marks, which will be crucial for determining the long-term efficacy. Second, the lack of a non-treatment controlled group limited our evaluation of the efficacy, analysis on biomarkers and cognition under an intention-to-treat framework are expected to better reveal the effect of lecanemab. Third, limited follow-up data on biomarkers and cognitive assessments have imposed restrictions on the generalizability of the findings. A comparison between the group with follow-up biomarkers and the group without follow-up biomarkers demonstrated that AEs, especially the incidence of ARIA, may affect their enthusiasm for biomarker monitoring. Finally, the western region of China was not included in this study, which limited the generalizability of the results.

In conclusion, our study provided a real-world analysis of lecanemab use in the Chinese population, offering valuable insights into its benefits. The rate of AEs was low and manageable, with favourable biomarker changes observed in the short term. Lecanemab administration was well-tolerated and safe in the moderate AD dementia population, providing a basis for future prospective studies. Further confirmation will require extended follow-up, and future large-scale studies are warranted.

Data availability

Anonymized participant data can be shared with qualified investigators on request to the corresponding author, from the date of publication onwards.

Funding

This study was supported by grants from the Science and Technology Innovation 2030 Major Projects (2022ZD0211600, 2021ZD0201806), the National Natural Science Foundation of China (82371188) and the National Key Research and Development Program of China (2023YFC3605400). Y.G. was funded by the National Natural Science Foundation of China (82501695), the National Postdoctoral Program for Innovative Talents (BX20240073), and China Postdoctoral Science Foundation (2024M760562).

Competing interests

The authors report no competing interests.

Supplementary material

Supplementary material is available at Brain online.