A Practical Approach to Selecting Patients for Amyloid-Targeting Treatments

Alzheimer disease (AD) is an incurable progressive neurodegenerative disease characterized by the accumulation of amyloid-β (Aβ) plaques and tau tangles, leading to cognitive decline and functional impairment.¹ New Food and Drug Administration (FDA)–approved amyloid-targeting treatments (ATTs)—lecanemab (Leqembi; Eisai, Nutley, NJ) and donanemab (Kisunla; Eli Lilly, Indianapolis, IN)—successfully target and clear cerebral amyloid plaques, modestly slow the progression of cognitive and functional decline, and positively affect several AD-associated biomarkers in people with early stages of decline.^2,3 The presence of amyloid-related imaging abnormalities (ARIAs), which manifest as edema (ARIA-E) or hemorrhage (ARIA-H) on brain MRI, is a relatively common and typically mild side effect but can be serious in rare instances. Selecting patients for ATTs involves identifying those most likely to benefit and at lower risk of developing side effects (particularly ARIA^2,4-6) to optimize the benefit-to-risk ratio. Clinical trials provide the essential framework for patient selection, but real-world use requires principled flexibility, inference, and judgment when considering patients and clinical contexts that are not well-represented in the trial data (Table). This review summarizes evidence-based practices related to patient selection, identifies potentially controversial areas with limited evidence, and shares our center’s (Mass General Brigham) approach to patients and contexts that fall outside an evidence-based practice.

Patient Selection Criteria
The primary task when selecting patients for ATTs is determining the likelihood of benefit and risk of side effects. This calculation is primarily informed by the key trials for each ATT (Clarity AD [NCT03887455] for lecanemab and TRAILBLAZER-ALZ2 [NCT04437511] for donanemab). The selection criteria used in these trials limit generalizability beyond the population studied, creating uncertainty about efficacy and safety in many real-world contexts. There may be clinical situations where there is reasonable likelihood of benefit despite the patient not strictly meeting criteria used in the trials. Appropriate use recommendations (AURs) have been developed for each ATT that provide guidance on how to approach selection for specific common contexts outside the trial population (eg, atypical AD syndromes).^6-8 Like many others, our center has adopted local institutional guidelines based on first principles (Table).⁹

Selecting patients for ATTs requires gathering a specific clinical data set to properly assess the risks and benefits of treatment, which supports decision-making as outlined within the published AURs for lecanemab and donanemab. The elements of this data set seek to create diagnostic confidence that AD is the primary etiology underlying the patient’s cognitive disorder, stage the cognitive illness severity, assess for known and potential risks of developing ARIAs, and identify comorbid conditions that could affect treatment safety, efficacy, adherence, or decision-making capacity (Table).

The most important trials informing ATT selection are Clarity AD¹ for lecanemab and TRAILBLAZER-ALZ2¹⁰ for donanemab.

Develop High Confidence that AD is Causing the Symptoms
A primary requirement for prescribing ATTs is a high clinical suspicion that AD neuropathology underlies the individual’s symptoms. Diagnostic confidence emerges from a well-characterized history of progressive cognitive and neuropsychiatric symptoms typical of AD, a suggestive cognitive profile, imaging findings with AD-pattern regional atrophy, and positive biomarker testing results. 

Clinical Syndrome. Clinician knowledge about the clinical–pathologic relationships in degenerative syndromes is essential to proper ATT patient selection. Typical, amnestic, multidomain cognitive decline with accompanying neuropsychiatric symptoms (eg, anxiety, irritability) in older individuals is the most common phenotype associated with neuropathologic AD and is easily recognized. Other nonamnestic syndromes, such as the logopenic variant of primary progressive aphasia (PPA) and posterior cortical atrophy, are also commonly associated with AD neuropathology, whereas others, such as the behavioral variant of frontotemporal dementia or other PPA variants, are not usually caused by AD. Positive AD biomarker testing alone, dissociated from the clinical syndrome, does not guarantee that AD is the likely cause of the symptoms, as this result is common in cognitively normal older adults and in syndromes where other neuropathologies are suspected to be driving symptoms (eg, dementia with Lewy bodies).

To characterize the clinical syndrome, it is important to screen for symptoms common in AD (eg, memory loss, word-finding difficulties, navigational difficulties) as well as those suggestive of alternative pathologies (eg, dream enactment, early parkinsonism, early personality changes). On cognitive testing, the pattern of cognitive deficits can support or refute a diagnosis of AD. Amnestic deficits on cognitive testing were required for Clarity AD trial eligibility.¹ The AURs extend eligibility to atypical AD variants based on their shared underlying pathophysiologic features with typical AD, and, in turn, the likelihood of similar benefit (Table). 

When there is uncertainty regarding whether AD pathology is a primary contributor to symptoms, additional diagnostic testing (eg, formal neuropsychologic testing, [¹⁸F]fluorodeoxyglucose positron emission tomography (FDG-PET) imaging, tau positron emission tomography (PET) imaging) may be helpful. These ancillary tests help clarify the pattern of abnormality at different levels, including the affected neuropsychologic processes and their associated functional neuroanatomy (neuropsychologic testing), regional metabolic functioning (FDG-PET), and the neuroanatomic distribution of tau (tau PET). Results suggestive of AD and alignment of symptoms increase diagnostic confidence that AD is the driving neuropathology. FDG-PET may yield patterns suggestive of non-AD pathologies, such as frontotemporal dementia (FTD) or Lewy body dementia, and therefore is useful when there is concern for alternative or copathologies. Tau PET imaging directly assesses the presence and distribution of tau pathology, providing both biomarker evidence of AD and brian regions affected by AD, which can help link symptoms with pathologic cause and identify whether AD pathology is driving symptoms. 

AD Biomarkers. Biomarker evidence of AD neuropathology is an essential inclusion criterion for ATT eligibility. Non-AD pathologies can cause syndromes that mimic AD and are commonly misdiagnosed as such on clinical grounds alone. Biomarkers can confirm the presence of AD neuropathology which is essential given amyloid’s status as the therapeutic target. In Clarity AD,¹ either amyloid PET via visual read or cerebrospinal fluid (CSF) AD biomarker testing were acceptable, whereas in TRAILBLAZER-ALZ2,¹⁰ amyloid PET with quantitative assessment was required. The AURs suggest confirming biomarker eligibility through either a positive visual read of a PET scan or CSF testing (Table).

Plasma biomarkers are an emerging option for determining the presence of AD pathology and may be comparable to CSF testing with lower cost and easier collection.^7,11,12 Examples include p-tau217, Aβ42/Aβ40, p-tau181, p-tau231, neurofilament light chain (NfL), and glial fibrillary acidic protein (GFAP).¹¹ In May 2025, the FDA approved its first plasma biomarker: a ratio of p-tau217 to Aβ1-42 assay.¹³ Insurance coverage for this assay remains uncertain. There is an emerging plasma biomarker of pathologic neurofibrillary tau, MTBR-tau243, which may correlate with symptoms and staging rather than AD pathology alone.¹⁴

Tau PET imaging is also emerging as a potential tool for refining ATT patient selection. Tau burden better reflects clinical symptoms and disease progression than does amyloid pathology alone.^13,15,16 Tau PET was obtained on all participants in TRAILBLAZER-ALZ2¹⁰ and a subset of participants in Clarity AD,¹ and demonstrated potential for predicting cognitive trajectories, with evidence suggesting that those with lower neocortical tau accumulation may experience slower progression and respond better to early intervention.^16,17 Although tau PET is not covered by most payors, incorporating this test could enhance risk stratification and optimize outcomes in early AD. In current practice, CSF and PET tests remain the standard for AD pathology detection.

Stage the Illness via Functional Assessment and Cognitive Testing
Clinical trials demonstrating effectiveness of ATTs are limited to early stages of AD including mild cognitive impairment (MCI) and mild dementia. FDA-approved ATTs have not been studied in more advanced AD^7,8 or preclinical AD, although trials for the latter are ongoing.¹⁸ AURs only include individuals with MCI or mild dementia; subjective cognitive impairment or moderate to severe dementia are exclusionary (Table). 

Staging is primarily determined by assessment of functional independence in everyday tasks. The Clinical Dementia Rating (CDR) scale was used in both key trials (ie, Clarity AD,¹ TRAILBLAZER-ALZ2¹⁰) and is the comprehensive standard for dementia staging. A global CDR score of 0.5 generally corresponds to a diagnosis of MCI, and 1.0 to mild dementia; CDR ≥2 suggests moderate dementia. However, the CDR is time-intensive and requires formal training which may limit its practicality in real-word practice settings. Validated alternative screening tools include the Functional Activities Questionnaire (FAQ)  and Quick Dementia Rating System (QDRS) which show strong correlations with the CDR^19,20 and are easier to administer. In our center, we use the FAQ and QDRS to screen functional status and a comprehensive CDR in cases with borderline staging (Table).

In addition to assessment of everyday functioning, an assessment of cognitive performance for the purpose of illness staging is helpful. In Clarity AD¹ and TRAILBLAZER-ALZ2,¹⁰ the Mini-Mental State Examination (MMSE) screening tool was used (in addition to the ADAS-Cog), with strict eligibility cutoffs of ≥22 (Clarity AD) and ≥20 (TRAILBLAZER-ALZ2). AURs include the Montreal Cognitive Assessment (MoCA) given its widespread use. Many multicenter studies validated a conversion relation between MMSE and MoCA, suggesting a MoCA score of 13 to 15 corresponds to an MMSE score of 22^8,21 and a MoCA score of 11 to 13 to an MMSE score of 20.⁷ In our center, we have adopted a guideline of MMSE score of 20 or a MoCA score of 14. 

Performance on cognitive screening assessments may be negatively affected by factors unrelated to an individual’s neuropathologic disease burden and interpretation requires nuance and contextualization. For example, longstanding or baseline cognitive deficits, language or cultural barriers, low educational attainment, test anxiety, and many other factors may lead to attainment of lower scores relative to disease burden. In addition, atypical AD variants, such as the logopenic variant of PPA or posterior cortical atrophy, may result in lower performance scores on cognitive testing relative to the clinical staging or pathologic burden. For this reason, in our center, the MMSE and MoCA thresholds serve as guidelines rather than strict exclusions, and each case below these scale’s cutoff scores is assessed for contributions from these other confounders. This is especially important in individuals where there is a mismatch between functional independence and cognitive performance. 

Assess ARIA Risk
Assessing each individual’s risk of developing ARIAs is an essential component of patient selection. ARIA has overlapping clinical and pathophysiologic features with cerebral amyloid angiopathy (CAA), including amyloid deposition within blood vessel walls accompanied by inflammatory activation. The presence of CAA risk factors or CAA neuroimaging features at baseline are hypothesized to correlate with the risk of developing ARIAs.

APOE Variants. A key driver of ARIA risk is the presence of the ε4 variant of the APOE gene. APOE ε4, in addition to its well-known role as a genetic risk factor for development of AD,⁸ increases the risk of developing ARIAs for individuals on ATTs.¹⁰ The APOE ε4-mediated risk of ARIA development is dose-dependent, with higher rates in homozygous individuals. For example, in Clarity AD,¹ ARIA-E incidence was 5.4% in noncarriers, 10.9% in heterozygotes, and 32.6% in homozygotes. Similar trends were observed with donanemab in TRAILBLAZER-ALZ2,¹⁰ although ARIA rates were higher. A later study using a modified dosing regimen of donanemab showed lower ARIA rates compared with standard dosing for participants with all APOE genotypes.²² ARIA-H also follows this dose-dependent pattern for both ATTs, with individuals with homozygosity showing the highest risk.^1,10 Thus, before determining eligibility, APOE genotyping should be obtained and the ARIA risk communicated to the patient and care partner. Some centers have elected not to treat individuals with ε4 homozygosity given the elevated risk of ARIAs. Our center treats individuals with ε4 homozygosity, but we have developed more stringent baseline MRI requirements. 

Baseline MRI Features. The clinical trials (ie, Clarity AD,¹ TRAILBLAZER-ALZ2¹⁰) also excluded people based on MRI findings considered to increase susceptibility to ARIAs. In both key trials, >4 microhemorrhages, any macrohemorrhage (ie, >1 cm), vasogenic edema, or evidence of CAA-related inflammation were exclusionary; Clarity AD1 also excluded participants with superficial siderosis (TRAILBLAZER-ALZ2 allowed a single focus of siderosis). Both studies excluded individuals with severe microvascular disease pattern changes on fluid-attenuated inversion recovery sequences. Fluid-attenuated inversion recovery white matter changes occur in CAA, which in principle could increase ARIA risk, but this finding is nonspecific and often reflects hypertensive or other small vessel pathologies. In our center, as in the AURs, we follow the shared criteria for exclusion and exclude individuals with baseline superficial siderosis. 

The key trials used gradient recalled echo (GRE/T2) sequences to evaluate for hemorrhage and 1.5T MRI magnet strength. However, alternative sequences (eg, susceptibility weighted imaging [SWI]) as well as higher magnet strengths (3T, 7T) can detect hemorrhage with increased sensitivity, which, if the same criteria are used, may produce a more stringent exclusion outcome than was used in the trials. Our center obtains both GRE and SWI, using the GRE as an inclusion criterion and the SWI to monitor for ARIA development during treatment. Given the critical role MRI plays in safety, individuals who cannot tolerate MRI even with light sedation are excluded.

Medications. Anticoagulants are known to increase hemorrhagic risk in CAA. In Clarity AD, participants on anticoagulation were not excluded but experienced a higher rate of macrohemorrhages.¹ Because of this, the AURs suggest excluding individuals on full-strength anticoagulation from receiving ATTs. It is reasonable to consider the likelihood of future anticoagulation needs, such as in individuals with atrial fibrillation or cryptogenic strokes, before initiating ATT. Aspirin and other antiplatelet treatments are not exclusionary. Dual antiplatelet therapy is of uncertain risk. For individuals receiving ATTs, we advise against the use of thrombolytics (eg, tissue-type plasminogen activator, tenecteplase). Although the absolute risk is uncertain, there are case reports of individuals on ATTs with fatal cerebral hemorrhage after thrombolytic treatment.²³ Given the potential shared vascular vulnerabilities between CAA and posterior reversible encephalopathy syndrome, medications associated with the latter should be used with caution.

Medical Conditions. Systemic medical conditions may affect the risk of developing ARIAs. In the trials (ie, Clarity AD, TRAILBLAZER-ALZ2) and AURs, poorly controlled bleeding disorders, especially those leading to a platelet count <50,000 PLT/μL or an international normalized ratio >1.5, were excluded due to the presumed increased risk of ARIA-H. Systemic autoimmune conditions were excluded given their potential interaction with immunologic mechanisms in ARIAs, with uncertain effects on safety and perhaps efficacy of treatment.

Identify Comorbid Conditions Potentially Affecting Safety, Efficacy, Adherence, or Decision-Making Ability
It is important to consider clinical, psychologic, and social context, as all these factors may affect safety, efficacy, adherence, understanding, and decision-making ability. Conditions that interfere with these treatment pillars should be considered as potentially exclusionary even when all other criteria are met. 

Comorbid Neurologic and Medical Conditions. People with suspected multiple pathologies, including diagnoses of probable Lewy body dementia or behavioral variant FTD, were excluded from the Clarity AD and TRAILBLAZER-ALZ2 trials. Our center has adopted this approach due to the known risks of ATT treatment but uncertain benefit in this context, especially when there is uncertainty regarding the driving underlying pathologic process. The risks and benefits of ATTs in other comorbid neurologic conditions (eg, idiopathic Parkinson disease, normal pressure hydrocephalus, small and stable meningiomas and cysts, nonsevere vascular cognitive impairment) especially when the AD component appears to be driving the symptomatology and prognosis is uncertain. Our center adjudicates these cases individually. AD in the context of Down syndrome was also exclusionary in the trials. 

In the Clarity AD and TRAILBLAZER-ALZ2 trials, a history of transient ischemic attack, symptomatic stroke, or generalized seizures within 12 months were exclusionary. Regarding stroke or transient ischemic attack, the risk of recurrent events is considered given safety concerns about thrombolytics which are commonly used in acute stroke. MRI-based criteria were used to help identify individuals with advanced cerebrovascular disease (Fazekas  score ≥3) or a significant history of stroke (≥2 lacunar infarcts or any major territorial strokes). A history of spontaneous intracranial hemorrhage is also exclusionary if not clearly explained by a prior condition expected to be self-limited without risk of recurrence (eg, traumatic brain injury).

Psychiatric conditions that interfere with treatment efficacy include diagnosed active psychosis, severe mood or personality disorders, and substance use disorders. In our center, we have adopted a criterion of stability for at least 6 months with possible additional monitoring requirements.

Comorbid systemic medical diagnoses, including malignancy, or cardiac, renal, hepatic, hematologic, endocrine, or immunologic conditions, may be exclusionary in some instances, especially when severe, when unstable, when ATT use may be affected by their treatment, and when the prognosis is uncertain. In our center, we review these cases individually and seek input from the individual’s medical teams with the goal of identifying these conditions’ effects on current and future treatment safety, efficacy, adherence, and decision-making capability. Individuals with autoimmune conditions requiring the use of immunologically active medications are generally excluded, mirroring the trial criteria. ATT-mediated amyloid clearance and ARIA are immune-mediated processes, and the potential effects on safety or efficacy in individuals with immunologic conditions are unknown. Our center attempts to identify individuals who may have lower-risk autoimmune diagnoses or features (eg, stable autoimmune thyroiditis) and engages them in shared decision-making about the uncertainties when considering whether to initiate treatment. 

Some AD-associated clinical features (eg, reduced awareness or insight regarding one’s cognitive deficits) have the potential to negatively affect care, specifically regarding capacity around shared decision-making, treatment adherence, and likelihood of reporting new symptoms while on treatment. Although reduced insight is not exclusionary, in our center we consider its impact in context and try to ensure close involvement of caregivers. The inability to provide informed consent is exclusionary.⁸

Care Partner Requirement. ATT protocols involve frequent infusions, MRIs, and follow-ups. This requires patients to have at least one reliable care partner who understands the treatment’s demands, risks, and benefits, and can coordinate the logistics of infusion appointments and visits (if the patient is unable to do so).^7,8 Without consistent care partner involvement, adherence to the treatment schedule may be compromised, making care partner involvement a crucial factor in both patient selection and ongoing management.

Challenges and Future Directions
The primary challenge in patient selection is identifying alignment of individual patients with evidence-based practice and recognizing when it is reasonable to extend eligibility to individuals outside the trial-based population and eligibility criteria. The limited racial and ethnic diversity in the Clarity AD¹ and TRAILBLAZER-ALZ2¹⁰ participant populations creates challenges when making inferences about safety and efficacy in people from underrepresented racial or ethnic groups. For all the inclusion and exclusion criteria discussed, there are borderline cases and potentially reasonable exceptions. In our center, when borderline cases or potentially reasonable exceptions are identified, members of our ATT team discuss the comorbidities in a case-by-case review and seek a consensus decision guided by input from the patient’s broader care team.^7,8,24-26 Patients offered ATTs are engaged in shared decision-making about the context of treatment, including the uncertainty regarding risks and benefits. Collecting real-world data derived from these cases will be essential to a better understanding of risks mediated by different factors and for assessing each individual’s risk, and will assist in developing a more personalized approach. 

ATTs are approved for individuals with early stages of AD. Individuals with “preclinical” AD who demonstrate biomarker evidence of AD neuropathologic changes but no illness-associated symptoms or cognitive deficits are not eligible for ATTs in clinic. Multiple trials designed to evaluate the efficacy of ATTs in this population are underway, including AHEAD 3–45 (NCT04468659)²⁷ and TRAILBLAZER-ALZ3 (NCT05026866),^10,28 which are studying the use of ATTs in prevention or delay of the emergence of AD symptoms or deficits.

As discussed in the AD Biomarkers section, emerging plasma biomarkers, such as p-tau217, Aβ42/Aβ40, and NfL, are promising, cost-effective alternatives to CSF testing. Tau PET imaging is also gaining attention for its ability to better reflect clinical symptoms and disease progression. However, despite these advances and the importance of biomarker testing, the essential component of ATT patient selection is a reliable clinical assessment to evaluate the likelihood that AD underlies the presenting symptoms. Training clinicians to develop expertise in making inferential diagnostic judgments is essential to achieving safe and effective outcomes, particularly in early stages of the disease.

Conclusion
ATTs represent a substantial advancement in the treatment of AD, but their optimal use depends on careful patient selection. This review highlights the complex and evolving criteria—including diagnostic, functional, cognitive, biomarker, genetic, and imaging requirements—that guide eligibility. Current frameworks offer a structured approach, but real-world implementation and emerging evidence suggest that patient selection for ATTs will likely become more dynamic, adaptive, and personalized in this postapproval phase, enhancing both safety and treatment outcomes.