Before the development of biomarkers, the 1984 NINCDS–ADRDA criteria clearly stated that “Diagnosis cannot be determined by laboratory tests,” and the diagnosis of AD relied entirely on clinical manifestations and the exclusion of other causes of dementia (McKhann et al., 1984). This framework reflected the limited understanding of AD pathophysiology at the time when in vivo markers of amyloid and tau pathology were not yet available.

In 2004, amyloid PET emerged as a groundbreaking imaging technology that enabled in vivo visualization of amyloid plaques and revolutionized AD research (Klunk et al., 2004). Among AD biomarkers, amyloid appears earliest and serves as an essential marker for diagnosis. However, the rate of Aβ deposition slows only a few years after reaching the positivity threshold and reaches a plateau by the MCI stage, making it challenging to use as a staging or monitoring marker (Jagust and Landau, 2021).

The first radiotracer developed, ¹¹C-labeled Pittsburgh compound B (¹¹C-PiB), binds with high sensitivity and specificity to fibrillar Aβ aggregates. A major limitation, however, is the short 20-min half-life of the ¹¹C isotope, restricting its use to research PET centers equipped with an on-site cyclotron. To overcome this, several ¹⁸F-labeled tracers with a longer half-life (−110 min) were developed, and ¹⁸F amyloid PET is now approved and used worldwide (Rabinovici et al., 2025).

Following the success of amyloid PET, tau PET capable of detecting NFTs was developed in 2013, marking another major milestone in AD imaging (Chien et al., 2013; Maruyama et al., 2013). By combining tau and amyloid PET, in vivo visualization of the two cardinal pathologies that define AD became possible for the first time, enabling investigation of the spatial and temporal interplay between Aβ and tau deposition.

In 2020, ¹⁸F-flortaucipir (FTP) became the first tau-selective tracer to receive approval from the U. S. Food and Drug Administration (FDA) for estimating the distribution and density of NFTs in patients with cognitive impairment being evaluated for AD. Although several other tau-targeted tracers have been developed, FTP remains the only one approved for clinical use. All tau tracers bind selectively to paired helical filament tau typical of AD, showing minimal-to-weak binding in non-AD tauopathies (Groot et al., 2022). Quantitatively, tau PET can estimate Braak stages V–VI and high ADNC with excellent sensitivity (92–100%) and moderate-to-high specificity (50–92%) (Fleisher et al., 2020).

Beyond AD, advances in tau imaging have expanded its application to non-AD tauopathies. The tracer originally known as PM-PBB3—now termed florzolotau (¹⁸F)—was developed as an extension of the earlier work (Maruyama et al., 2013) and has demonstrated high-contrast detection of progressive supranuclear palsy (PSP) and corticobasal degeneration (CBD) tau pathology (Tagai et al., 2021). This pan-tau imaging capability represents an important step toward differential diagnosis among tauopathies.

Early studies in the 1990s, using sensitive sandwich ELISA methods, demonstrated that cerebrospinal fluid (CSF) total tau (t-tau)—a soluble N-terminal fragment derived from the axonal tau protein rather than the filament core of NFTs—was elevated in clinically diagnosed AD, whereas Aβ42, the principal peptide forming amyloid plaques, was decreased (Vandermeeren et al., 1993; Arai et al., 1995; Blennow et al., 1995; Motter et al., 1995). Initial work with phosphorylation-specific antibodies (e.g., AT8; Ser199–202) showed limited CSF reactivity, but subsequent assays targeting epitopes such as T181, S199, T205, T217, and T231 confirmed consistent increases in phosphorylated tau (p-tau) in AD (Xia et al., 2021; Gobom et al., 2022). These findings established CSF biomarkers as reliable tools for in vivo diagnosis of AD (Olsson et al., 2016; Iaccarino et al., 2023). Although lumbar puncture is generally safe and well tolerated, its perceived invasiveness, need for specialized staff, and contraindications (e.g., anticoagulant use) have limited broader use (Schindler and Atri, 2023). Consequently, attention has shifted toward blood-based biomarkers that are easier to collect and more scalable in clinical settings.

Early studies also reported increases in plasma Aβ in AD during the 1990s, but limited assay sensitivity and reproducibility hindered clinical application (Mehta et al., 2000; Hansson et al., 2010). This changed with the development of immunoprecipitation–mass spectrometry (IP-MS), which enabled accurate measurement of plasma Aβ42/40 ratios and demonstrated a strong correlation with abnormal brain amyloid burden (Ovod et al., 2017; Nakamura et al., 2018). Subsequently, plasma p-tau biomarkers showed strong concordance with amyloid and tau PET as well as CSF biomarkers, reflecting ADNC with high diagnostic accuracy (Janelidze et al., 2020; Karikari et al., 2020; Palmqvist et al., 2020; Thijssen et al., 2020; Janelidze et al., 2021). Among them, p-tau217 performs comparably to CSF measures in classifying amyloid and tau PET status, markedly increasing the clinical utility of blood-based biomarkers (Barthélemy et al., 2024).

Nearly three decades after the 1984 NINCDS–ADRDA criteria, the 2011 NIA–AA guidelines revised the clinical diagnostic framework for dementia due to AD (McKhann et al., 2011). The updated criteria remained primarily clinical, defining the dementia syndrome based on cognitive and functional findings. However, for the first time, CSF and amyloid PET biomarkers were incorporated as supportive evidence of the underlying pathophysiology, although routine biomarker testing was not yet recommended for clinical use. Structural MRI markers of neurodegeneration, including medial temporal lobe atrophy, were also recognized as supportive features of AD-related neurodegeneration in research settings. In a landmark conceptual shift, companion articles proposed viewing AD as a continuum of clinical and biological processes (Albert et al., 2011; Sperling et al., 2011). Whereas the term Alzheimer’s disease had previously referred only to the dementia stage, the revised framework expanded the construct to include MCI due to AD and, for research purposes, asymptomatic individuals with positive amyloid or CSF biomarkers—termed preclinical AD. This redefinition established the foundation for the biological staging of AD that would later be formalized in the 2018 and 2024 NIA–AA criteria.

Driven by advances in biomarkers and by recognition of the limitations of symptom-based definitions in a disease that unfolds along a continuum, the 2018 NIA–AA research framework redefined AD in biological terms for research purposes (Jack et al., 2018). The framework proposed that “Alzheimer’s disease is defined by its underlying pathologic processes, which can be documented by postmortem examination or in vivo by biomarkers. The diagnosis is not based on the clinical consequences of the disease (i.e., symptoms or signs)”. Importantly, neurodegeneration markers, including structural MRI measures, were explicitly repositioned as indicators of disease severity rather than defining features of AD pathology. The framework had three main objectives: (i) to standardize research methodology, (ii) to refine participant selection for disease-modifying therapies (DMTs) trials, and (iii) to improve understanding of disease etiology. Following the AT(N) classification system, AD was defined biologically as the presence of both A (aggregated amyloid-β or its associated pathologic state: CSF Aβ42 or Aβ42/Aβ40 ratio, or amyloid PET) and T (aggregated tau (NFTs) or its associated pathologic state: CSF phosphorylated tau or tau PET) positivity. In the 2011 NIA–AA guidelines, CSF p-tau and t-tau had been regarded together as markers of neuronal injury. However, since p-tau rises more specifically in AD, whereas t-tau can be elevated in other neurological disorders, the 2018 framework designated p-tau as the T marker and t-tau as the N (neurodegeneration) marker. At that time, “p-tau” referred primarily to p-tau181, the most validated phospho-epitope. Because this biological definition had not yet been validated in large longitudinal cohorts or randomized clinical trials, the authors explicitly restricted its use to a research framework, emphasizing that “it is premature and inappropriate to apply this research framework in general medical practice”.

Over the following years, rapid progress in biomarker technology—especially the development of highly accurate plasma p-tau assays and tau PET imaging—narrowed the gap between research and clinical application. The approval of anti-amyloid therapies further underscored the need for standardized, biologically grounded diagnostic and staging criteria that could guide both care and research.

The 2024 NIA–AA criteria provide a biology-based framework for the diagnosis and staging of AD, aiming to bridge research and clinical practice (Jack et al., 2024). The revision was undertaken for three main reasons: (i) with the approval of DMTs, standardized diagnostic and staging systems are needed across clinical, research, and industry settings; (ii) the emergence of blood-based biomarkers has made biological diagnosis increasingly accessible; and (iii) accumulating evidence supports the partial interchangeability of different biomarker modalities. The revised framework is grounded in several fundamental principles: AD is defined by its biology rather than by its symptoms; biomarker evidence of AD neuropathology is sufficient for diagnosis; AD exists on a continuum that begins asymptomatically and progresses biologically, with clinical symptoms emerging later but not required for diagnosis; and similar clinical syndromes may result from non-AD causes, meaning that symptoms alone are not diagnostic. Nonetheless, these criteria are not intended as clinical practice guidelines. Because biomarker testing is not universally accessible, the document is not designed as a step-by-step protocol for all care settings, nor does it prescribe specific treatment pathways.

In α-synucleinopathies such as Parkinson’s disease (PD) and dementia with Lewy bodies (DLB), it has been proposed, analogous to AD, to define disease entities biologically rather than by clinical syndromes (Simuni et al., 2024). This paradigm shift was made possible by the advent of a robust biomarker, the CSF α-synuclein seed amplification assay (SAA), which exploits the prion like propagation of α-synuclein to enable biological classification (Fairfoul et al., 2016; Shahnawaz et al., 2017; Grossauer et al., 2023). The availability of such a biological marker has facilitated the identification of α-synuclein co-pathology within AD using multimodal biomarker profiles. Approximately half of AD cases reportedly harbor α-synuclein (Lewy body) pathology. Compared with “pure” AD, mixed AD/Lewy body cases more frequently present with DLB like clinical features and show faster cognitive decline and higher mortality than either pure AD or pure DLB (Thal et al., 2025). Physiologically, α-synuclein may interact with Alzheimer’s pathology to accelerate disease progression, potentially accounting for inter-individual variability in the clinical course (Gawor et al., 2024). Moreover, α-synucleinopathy inferred by SAA influences cognitive function independently of AD pathology, both cross sectionally and longitudinally, thereby contributing substantially to discrepancies between biological and clinical staging when present as a co-pathology (Palmqvist et al., 2023; Quadalti et al., 2023). Beyond CSF-based SAA, several other α-synuclein biomarkers, including plasma SAA, RT QuIC assays, and PET tracers under development, have been reported, which may further refine the in vivo assessment of α-synuclein co-pathology.

Although the 2024 NIA-AA criteria list vascular brain injury (VBI) as a non-AD co-pathology, it does not provide a precise, unified definition. Indeed, multiple definitions exist for vascular cognitive impairment (VCI) (You et al., 2024; Sachdev et al., 2025; Smith et al., 2025; Swartz et al., 2025). In this review, the term VBI broadly encompasses vascular pathology with a spectrum of clinical and subclinical presentations, including cortical and subcortical infarcts, strategic infarcts, small-vessel disease (SVD) with white matter lesions, lacunes, and intracerebral hemorrhages. Despite terminological heterogeneity that complicates epidemiological comparisons, VCI is widely recognized as the second most common cause of dementia, accounting for approximately 20–40% of all cases (Lobo et al., 2000; Kalaria et al., 2008; Chang Wong and Chang Chui, 2022; Rundek et al., 2022; Tuominen et al., 2025).

VBI and AD share aging as a major risk factor and have numerous additional overlapping risk factors (Sato and Morishita, 2015). According to autopsy studies, vascular pathology is about four times more common in AD brains that show mixed pathology than in those with AD pathology alone (Medina-Rioja et al., 2024). Moreover, vascular risk factor burden and AD pathophysiology progress independently yet synergistically, together accelerating neurodegeneration and cognitive decline (Ferrari-Souza et al., 2024). In addition to its relationship with AD, VBI may also interact with TDP-43–related pathology described later in this review, suggesting shared vascular and metabolic vulnerabilities (Grothe et al., 2023; Nelson et al., 2024). Further accumulation of clinicopathological and biomarker evidence will be necessary to clarify whether these associations are causal or simply coexistent.

Among the causes of VCI, SVD represents the major contributor, accounting for approximately 50–70% of cases (Cheng et al., 2024). SVD includes heterogeneous small-vessel pathologies such as arteriolosclerosis and cerebral amyloid angiopathy (CAA). On MRI, SVD is characterized by a constellation of imaging features, including white matter hyperintensities (WMH) and other imaging findings, rather than WMH representing a stand-alone etiologic category. Of note, WMH are increasingly recognized as pathophysiologically heterogeneous lesions and may reflect not only SVD-related vascular injury but also non-vascular mechanisms, including AD-related neurodegenerative processes (Garnier-Crussard et al., 2023). Therefore, Table 1 clarifies the hierarchical relationship among these commonly conflated terms (SVD: disease construct; WMH: imaging marker; and CAA: underlying small-vessel pathology). Hallmarks of SVD include recent small subcortical infarcts, lacunes, white matter hyperintensities (WMH), enlarged perivascular spaces, cerebral microbleeds, cortical superficial siderosis, brain atrophy, and cortical cerebral microinfarcts (Duering et al., 2023). As an AD co-pathology, a higher WMH burden particularly increases the risk of incident dementia and AD (Debette et al., 2019; Duering et al., 2023). Both Aβ pathology and WMH independently impair cognitive performance and synergistically exacerbate future cognitive decline (Ali et al., 2023). Beyond their vascular origin, recent evidence indicates that a subset of WMH in AD may arise from AD-related processes such as tau-mediated axonal degeneration, amyloid toxicity, and neuroinflammation rather than ischemia, highlighting the heterogeneous nature of WMH pathophysiology (Garnier-Crussard et al., 2023). Recognizing this heterogeneity will be essential for interpreting WMH as biomarkers within the AD continuum and for differentiating vascular from degenerative mechanisms in mixed dementias.

Unlike other forms of VBI, cerebral amyloid angiopathy (CAA) is etiologically related to impaired Aβ processing rather than to traditional systemic vascular risk factors (Wang et al., 2024). Autopsy evidence indicates that 80–100% of AD patient’s exhibit CAA-like amyloid deposition in intracranial vessels, whereas 40–60% of patients with CAA-associated intracerebral hemorrhage show AD-like neuropathological changes (Wang et al., 2024). Pathophysiologically, CAA involves aggregation of Aβ within vessel walls, resulting in vessel fragility (Greenberg et al., 2020). CAA contributes to cognitive decline, although the mechanisms remain under debate. Competing models propose that cognitive impairment arises either directly from CAA-related vascular dysfunction or indirectly through interactions with AD pathology—particularly NFTs (Boyle et al., 2015; Rabin et al., 2022; Godrich et al., 2024; Sin et al., 2024). Clinically, an important complication of CAA is lobar intracerebral hemorrhage, which carries a mortality rate of approximately 10–40% (Wang et al., 2024). CAA is also a recognized risk factor for amyloid-related imaging abnormalities (ARIA), as emphasized in anti-amyloid therapy labels, underscoring the clinical importance of its detection (Hampel et al., 2023; Eisai Inc, 2025; Eli Lilly and Company, 2025b).

In the 2024 NIA-AA criteria, the most definitive vascular biomarkers are anatomic MRI or CT evidence of macroscopic cerebral infarctions; however, no single summary measure yet exists to comprehensively quantify vascular pathology (Jack et al., 2024). Diffusion-weighted imaging is sensitive to SVD-related damage but lacks specificity, and WMH are not unique to SVD (Duering et al., 2023; Silbert et al., 2024). We and others have recently demonstrated that plasma PlGF levels correlate with WMH volume, underscoring its potential role as a surrogate marker for SVD (Igarashi et al., 2025). Continued efforts are underway to standardize and validate new biomarkers and imaging modalities for vascular co-pathologies (Duering et al., 2023).

Limbic-predominant age-related TDP-43 encephalopathy (LATE) is a clinicopathological entity characterized neuropathologically by a stereotypical pattern of TDP-43 proteinopathy, termed LATE neuropathologic change (LATE-NC), and clinically by a slowly progressive amnestic cognitive disorder in older adults that is often difficult to distinguish from AD (Nelson et al., 2019).

It has become increasingly recognized that AD represents only one of several neuropathological diseases associated with amnestic cognitive impairment. In particular, older individuals, especially the oldest-old, may exhibit cognitive decline out of proportion to the severity of ADNC (Schneider et al., 2007; Savva et al., 2009; Nelson et al., 2012). Among the major contributors identified was TDP-43 proteinopathy, now acknowledged as a distinct pathogenic mechanism in late life (Nelson et al., 2013; Josephs et al., 2015). In 2019, a formal consensus first established neuropathologic diagnostic criteria and staging for LATE (Nelson et al., 2019).

Because LATE-NC increases with age, estimates vary across community- or population-based autopsy cohorts with different age-at-death profiles. In a combined analysis of 13 such cohorts, the proportion with LATE-NC ranged from 11.1 to 67.7% (overall ≈39.4%), and cohort mean ages at death spanned 72.2–97.2 years (Nelson et al., 2022). In a strictly population-based study restricted to the oldest-old (Vantaa 85+; all residents ≥85 years), any LATE-NC stage was present in 64.1% of individuals—highlighting how common LATE is at very advanced ages (Mikhailenko et al., 2025). Note that these are autopsy-based frequencies and “do not represent projected population prevalence”. LATE-NC frequently co-occurs with ADNC, affecting more than 50% of those with frequent neuritic plaques (Nelson et al., 2022; Mikhailenko et al., 2025). Such co-pathologies markedly accelerate cognitive decline (Boyle et al., 2017; Katsumata et al., 2020; Montine et al., 2022; Smirnov et al., 2022) and are also associated with memory decline independent of ADNC, which has motivated efforts to develop in vivo biomarkers (Nelson et al., 2022; Mikhailenko et al., 2025). With the clinical rollout of anti-amyloid therapies, accurate in vivo diagnosis of LATE has become increasingly important for interpreting DMTs outcomes. However, validated biomarkers for TDP-43 or LATE-NC are currently lacking. Accordingly, proposed clinical diagnostic criteria for LATE incorporate clinical features, imaging findings, and AD biomarker status (Wolk et al., 2025). The criteria suggest that LATE should be considered the primary driver when both core clinical syndrome and a required imaging finding are present. Clinically, this refers to a progressive amnestic syndrome characterized by prominent episodic memory loss, often described as temporo-limbic amnesia, with relative preservation of other cognitive domains, and on imaging to marked hippocampal atrophy that is disproportionate to global atrophy. When LATE is suspected as the primary pathology, a negative amyloid biomarker supports a diagnosis of probable LATE. If amyloid is positive, tau biomarkers are evaluated: a negative tau result indicates possible LATE, whereas a positive tau result necessitates a comprehensive assessment using clinical trajectories and MRI, tau PET, or FDG-PET to determine whether the findings can be explained by AD alone. If not, the diagnosis is possible LATE with AD. In essence, these criteria emphasize the exclusion of findings typical of AD. They serve as an initial framework, and further validation together with the development of specific biomarkers is anticipated.

In research settings, in vivo biomarkers for TDP-43 proteinopathies are under active development, but no disease-specific biomarker for LATE-NC has yet been established. Fluid-based assays face substantial challenges, as TDP-43 is a ubiquitously expressed protein and most currently available antibodies fail to reliably distinguish pathological from physiological species in CSF or plasma. Consequently, measured TDP-43 levels may reflect diverse systemic or neurological injuries rather than LATE-specific pathology. Moreover, the majority of CSF and plasma TDP-43 studies to date have focused on amyotrophic lateral sclerosis or frontotemporal lobar degeneration, limiting their applicability to LATE-NC.

A recent study using a highly sensitive plasma assay demonstrated an association between plasma TDP-43 levels and autopsy-confirmed advanced LATE-NC, particularly in individuals with comorbid AD pathology; however, its ability to detect isolated LATE-NC was less clear, highlighting the current limitations of fluid biomarkers for this condition (Wang et al., 2025b). Although several PET ligands targeting TDP-43 pathology have been proposed at a preclinical or exploratory stage (Vokali et al., 2025), none have yet demonstrated sufficient specificity or validation for clinical in vivo use. Consequently, direct PET-based assessment of TDP-43 pathology in LATE-NC remains unavailable.

In the absence of disease-specific biomarkers, several imaging features have been proposed as supportive indicators of LATE. These include medial temporal and hippocampal atrophy on structural MRI and an elevated inferior temporal-to–medial temporal lobe ratio on FDG-PET, which may help distinguish LATE-related patterns from typical AD trajectories (López-Carbonero et al., 2024).

Finally, under the 2024 NIA-AA criteria, multimodal AT₁T₂NISV biomarker profiles can serve as useful indirect indicators of co-pathologies that lack disease-specific biomarkers, such as LATE. In particular, a biomarker profile showing neurodegeneration without tau aggregation (T₂ − N + profile, also referred to as a T₂N mismatch) suggests the presence of non-AD injury.

Individuals whose clinical stage is better than expected for their biological stage are thought to possess exceptional resilience or resistance. Numerous studies have explored the factors that contribute to this phenomenon. However, because different research fields have used overlapping terms with varying definitions, confusion can arise; therefore, we first clarify terminology. Here, terms are organized according to the framework proposed by the National Institute on Aging’s Collaboratory (Stern et al., 2023). Resilience is used as an umbrella term defined as “a general term that subsumes any concept that relates to the capacity of the brain to maintain cognition and function with aging and disease.” Within this overarching construct, the framework provides operational definitions for cognitive reserve (CR), brain maintenance (BM), and brain reserve (BR). This concept is illustrated in Figure 2, which integrates biological and clinical staging under the 2024 NIA-AA criteria and depicts how co-pathologies can accelerate, and resilience can delay, clinical progression despite equivalent levels of biological pathology.

Building on this conceptual framework, CR refers to individuals who maintain better cognitive performance despite a similar degree of neuropathology. In empirical studies, CR is commonly operationalized using sociobehavioral proxies such as early age IQ, educational attainment, occupational complexity, and cognitively stimulating leisure activities, which modulate the relationship between biomarker burden and clinical expression (Stern, 2012).

BM refers to those whose neuropathology and brain atrophy progress more slowly than that of their peers over time. Genetics and lifestyle can impact BM. For example, APOE genotype modulates vulnerability across the AT(N) cascade, with ε2 conferring relative protection and ε4 increasing susceptibility to amyloid and tau pathology (Tzioras et al., 2019).

BR refers to those who begin life with greater intracranial or neuronal resources and therefore manifest symptoms later in life. Proxies to estimate BR include intracranial volume, head circumference, specific patterns of gray matter volume, and cortical surface area (Stern et al., 2020).

By contrast, resistance is defined as “absence or lower than expected levels of AD pathology” (Arenaza-Urquijo and Vemuri, 2020). Accordingly, within the NIA-AA integrated staging system, individuals who present with a better clinical stage than expected for their biological stage can be understood as demonstrating CR within the broader construct of resilience.

Many studies have reported that sociobehavioral factors such as education, occupation, and leisure activities, as well as genetic factors such as APOE, contribute to CR (Jagust and Mormino, 2011; Vemuri et al., 2014; Pappalettera et al., 2024). However, because terminology and measurement approaches remain inconsistent, high-quality and reproducible evidence has been limited, emphasizing the need for greater standardization in future research (Pappalettera et al., 2024; Mutel et al., 2025).

Across multiple cohorts applying the 2024 NIA-AA criteria, concordance between biological and clinical staging is estimated at only about 30–50%, demonstrating substantial heterogeneity across the AD continuum (Lu et al., 2024; Mendes et al., 2025; Pichet Binette et al., 2025). This heterogeneity is driven by both co-pathology and resilience. Among these, discordance in which the clinical stage exceeds the biological stage is more common, underscoring the impact of co-pathologies on symptom expression. Conversely, fewer cases show a more advanced biological than clinical stage, likely because tau PET is strongly linked to cognitive decline and clinical progression. These patterns vary depending on cohort composition and the operational definitions used for tau PET–based staging, which are intentionally broad within the 2024 NIA-AA criteria (Mendes et al., 2025; Pichet Binette et al., 2025).

As illustrated in Figure 2, individuals with comparable AT(N) biomarker profiles can occupy markedly different clinical stages, highlighting how resilience-related factors and co-pathologies shape real-world disease trajectories beyond biomarker-defined classifications.

Taken together, these findings indicate that discordance between biological and clinical staging is not an exception but a common feature of AD. Such heterogeneity highlights the limitations of relying solely on biomarker-defined stages to predict clinical outcomes and highlights the importance of incorporating modifiers of reserve and resilience when interpreting disease trajectories. In this context, resilience-related factors provide a conceptual framework for understanding why individuals with comparable AT(N) profiles may follow divergent clinical courses.

In the United States, lecanemab received accelerated approval from FDA in 2023 and, following the results of the phase 3 trial, became the first DMT for AD to receive full FDA approval (U.S. Food and Drug Administration, 2023b; van Dyck et al., 2023; Eisai Inc, 2025). After full approval, the therapy was also covered by Medicare, enabled broad clinical implementation (Centers for Medicare and Medicaid Services, 2023b). The indication is consistent across FDA labeling and Medicare coverage and aligns with the inclusion criteria used in clinical trials, specifically targeting patients in the MCI or mild dementia stage. Before initiating therapy, the FDA labeling specifies that “the presence of amyloid beta pathology must be confirmed prior to initiating treatment,” and Medicare likewise states that “your doctor or other health care provider must confirm you have beta-amyloid plaques consistent with Alzheimer’s disease,” without specifying a particular diagnostic modality (U.S. Food and Drug Administration, 2023b; Centers for Medicare and Medicaid Services, 2025). During treatment, brain MRI is required to monitor for ARIA.

Japan followed the United States in approving lecanemab and listing it for National Health Insurance reimbursement in 2023 (Eisai Co., Ltd, 2023). Japan’s optimal-use guideline explicitly requires confirmation of amyloid pathology by amyloid PET or CSF (Ministry of Health, LAW, 2023a). MCI or mild dementia due to AD, defined using both the Mini-Mental State Examination (MMSE) and the Clinical Dementia Rating (CDR 0.5–1). For safety monitoring, in addition to the FDA-recommended MRI scans before the 5th, 7th, and 14th doses, MRI every 6 months thereafter is recommended.

In the European Union (EU), lecanemab initially received a negative opinion but was later authorized by the European Commission in April 2025 following re-examination. However, because of the risk of ARIA, the indication is restricted to patients with early-stage AD who carry either zero (non-carriers) or one copy (heterozygous carriers) of the APOE ε4 allele (European Medicines Agency, 2025b).

Donanemab is a novel anti-amyloid monoclonal antibody designed for administration over a fixed treatment period (Sims et al., 2023). Whereas lecanemab primarily targets soluble Aβ aggregates (protofibrils), donanemab targets pyroglutamate-modified amyloid beta at the third amino acid position (N3pG-Aβ), which is believed to exist only within insoluble amyloid plaques.

Donanemab was approved by FDA in 2024 and subsequently covered by Medicare. Its indication mirrors that of lecanemab, targeting patients with MCI or mild dementia due to AD, and includes requirements for regular MRI monitoring (Centers for Medicare and Medicaid Services, 2022; Eli Lilly and Company, 2025b). The therapy was listed for National Health Insurance reimbursement in Japan in 2024 (Eli Lilly Japan KK, 2024; Ministry of Health, LAWJ 2025b). In the EU, donanemab received authorization from the European Commission in September 2025, with its indication similarly restricted to APOE ε4 non-carriers or heterozygous carriers (European Medicines Agency, 2025a). Pricing and reimbursement are determined separately by each member state’s national health authorities.

Lecanemab and donanemab are both anti-amyloid monoclonal antibodies approved for early AD. The key distinction lies in their targets: lecanemab binds soluble amyloid beta protofibrils, whereas donanemab recognizes pyroglutamate-modified amyloid beta within insoluble plaques. Both require confirmation of amyloid pathology and regular MRI monitoring for ARIA, but donanemab is designed for a fixed-duration course, while lecanemab is administered continuously. Because ARIA risk is higher in APOE ε4 homozygotes, the EU restricts treatment to non-carriers and heterozygotes. Overall, both agents mark the beginning of biomarker-guided precision therapy in AD.

It should be noted that no head-to-head clinical trials directly comparing lecanemab and donanemab have been conducted to date. As a result, differences in efficacy and safety between these agents are inferred from separate trials with distinct designs, inclusion criteria, and outcome measures, and should be interpreted with caution. Moreover, post-approval experience has highlighted that the incidence, management, and practical impact of ARIA may differ in real-world settings compared with rates reported in pivotal trials. For example, early real-world data from Japan indicate that while treatment discontinuation due to ARIA has been relatively infrequent, substantial logistical and infrastructural challenges remain, including MRI capacity, infusion resources, and access to APOE genotyping, highlighting the complexity of implementing disease-modifying therapies beyond regulatory trial settings (Sato et al., 2018).