Eligibility for Anti-Amyloid Therapies in Biomarker-Confirmed Atypical Alzheimer’s Disease Phenotypes

Abstract

Background and Objectives:

Clinical trials of anti-amyloid therapies (AAT) for Alzheimer’s disease (AD) primarily enrolled patients with mildly symptomatic, amnestic predominant presentations. The applicability of eligibility criteria to atypical AD phenotypes, including posterior cortical atrophy (PCA), logopenic variant primary progressive aphasia (lvPPA), dysexecutive AD (dAD) and corticobasal syndrome (CBS-AD), is unknown.

Methods:

We conducted a retrospective eligibility analysis of atypical AD patients evaluated at Mayo Clinic. Theoretical eligibility for AAT was assessed at initial clinical evaluation by applying inclusion and exclusion criteria from landmark clinical trials (CLARITY-AD and TRAILBLAZER-ALZ2) and appropriate use criteria for lecanemab and donanemab. Eligibility percentages and reasons for exclusion were compared across phenotypes.

Results:

The cohort included 184 patients (61.4% female) with biomarker-confirmed atypical AD: PCA (n=98, 53.3%), lvPPA (n=42, 22.8%), dAD (n=37, 20.1%) and CBS-AD (n=7, 3.8%). Age at onset (p=0.039) and presentation (p<0.001) differed, with lvPPA patients oldest (median age at onset, presentation: 63.1, 66.9 years) and dAD patients youngest (onset, presentation: 55.1, 57.3). Functional impairment differed by phenotype (p = 0.005), with lvPPA more often diagnosed at the MCI/very mild stage (CDR 0.5; 71.4%), whereas PCA and dAD more frequently presented with mild dementia, although time from symptom onset to diagnosis did not differ (p = 0.634). MMSE scores differed (p=0.036), with lower scores in PCA and lvPPA (median 21 and 21, respectively) compared to CBS-AD (median 27). Depending on the eligibility framework applied, 70-85% of patients would not meet treatment criteria. Bedside cognitive thresholds were the primary drivers of ineligibility (50-67% of exclusions), despite most patients having early symptomatic disease (global CDR 0.5-1; 82%). Imaging-based exclusions (22-27%) and severity thresholds (moderate or severe dementia, 19-23%) were also frequent. Reasons for ineligibility were similar across phenotypes.

Discussion:

Most patients with atypical AD would not meet eligibility criteria for AAT, typically due to MMSE-based exclusion rather than measures of cognitive function. Eligibility rates are broadly similar across atypical phenotypes. These findings highlight the need for phenotype-sensitive staging and patient selection for treatment in atypical AD.

Introduction

Anti-amyloid monoclonal antibody (AAT) trials, including CLARITY-AD ¹ (lecanemab) and TRAILBLAZER-ALZ2 ² (donanemab), primarily enrolled individuals with typical, amnestic predominant presentations of AD, with inclusion restricted to patients with mild cognitive impairment (MCI) or mild dementia, and endpoints optimized for memory-predominant impairment ^3,4. Accordingly, patients with atypical AD phenotypes were largely excluded from clinical trials, including individuals with posterior cortical atrophy (PCA), logopenic variant primary progressive aphasia (lvPPA), dysexecutive AD (dAD), behavioral variant AD and corticobasal syndrome due to AD (CBS-AD). Evidence regarding eligibility, efficacy and safety of AAT in these populations is limited ^3,5.

Although the underlying neuropathological processes are fundamentally similar in patients with typical and atypical AD phenotypes ⁶, patients with atypical phenotypes commonly demonstrate greater tau burden and neurodegeneration at first presentation, with the neocortical distribution of neuropathology mirroring their clinical syndrome, and with relative sparing of the hippocampus ^7-10. These presentations are also more common in younger patients (<65 years)⁷. The combination of earlier age at symptomatic onset and non-amnestic presenting features contributes to delayed recognition and referral and is associated with higher rates of misdiagnosis and diagnostic delay, as their symptoms may be attributed to psychiatric illness or life stressors, and standard neuropsychological assessments may not adequately capture their deficits ^7,11-13. In a neuropathologically confirmed cohort, over half of young-onset atypical cases were initially misdiagnosed, compared to a smaller proportion of typical cases ⁷. As a result, these patients may be less likely to meet timely clinical eligibility criteria for AAT. Individuals with atypical AD variants also remain underrepresented in large, trial-oriented datasets and therapeutic clinical trial infrastructures, which often emphasize memory-predominant impairment and age-based cutoffs ¹⁴. As a result, eligibility patterns derived from typical amnestic AD populations may not generalize to these phenotypes.

We systematically evaluated treatment eligibility using the inclusion and exclusion criteria from the pivotal clinical trials (CLARITY AD, TRAILBLAZER-ALZ2) as well as appropriate use recommendations for lecanemab ¹⁵ and donanemab ¹⁶ in four biomarker-confirmed atypical AD cohorts. Our objective was to determine the proportion of biomarker-confirmed atypical AD patients at their presentation who would meet eligibility criteria for AAT, in order to inform clinical decision-making in a population where clinical judgment currently plays a central role ¹⁷.

Methods

Standard Protocol Approvals, Registrations, and Patient Consents

Individuals included in the study had previously provided informed consent for research use of their medical records. Of the 184 participants, 112 were enrolled in longitudinal research studies and provided written informed consent as part of those protocols; the remaining participants had signed authorization permitting use of their medical records for research. The study was reviewed by expedited review procedures of the Mayo Clinic Institutional Review Board (IRB) and was deemed minimal risk (Study ID: 25-011269).

Study Participants

This was a retrospective analysis of patients evaluated primarily prior to the regulatory approval and clinical implementation of AAT. Most participants (178/184) were assessed before AAT implementation (February 2000–July 2023) and therefore were not prospectively evaluated for treatment eligibility. As such, data completeness reflects routine clinical practice rather than standardized screening for AAT. Only six participants were evaluated after AAT implementation at Mayo Clinic (December 2023–July 2024).

We conducted a retrospective eligibility analysis using pooled data from previously established, well-characterized cohorts comprising patients with biomarker-confirmed atypical AD variants. These included individuals with PCA ¹¹, diagnosed according to consensus criteria ¹⁸; CBS-AD ¹⁹, meeting established CBS criteria ²⁰; and dAD ²¹, characterized as previously described ²². Patients with lvPPA were drawn from two cohorts ^23,24 and met established consensus criteria ²⁵.

Criteria assessments

Trial inclusion and exclusion criteria were obtained from the original phase 3 trial publications and corresponding appropriate use guidelines (Table 1). Eligibility was determined by independently applying inclusion and exclusion criteria for each framework to each participant, rather than using a stepwise screening approach, based on clinical data available at presentation or diagnosis. Records were reviewed by a single reviewer (D.S.) for eligibility assessment.

Table 1. Comparative Eligibility Criteria: Lecanemab vs Donanemab, Trial and Appropriate Use Recommendation.

Category	Lecanemab CLA RITY-AD TRIAL	Lecanemab Appr opriate Use Recommendations	Donanemab TRAILBLAZER- ALZ2	Donanemab Appropriate Use Recommendations
Age	50–90 yrs	50–90 yrs, clinical judgment allowed	60–85 yrs	60–85 yrs, clinical judgment allowed
Clinical Duration	≥1 year	-	≥6 months	-
Clinical Stage	MCI or mild AD dementia; CDR 0.5–1.0; CDR Memory Box ≥0.5	MCI or mild AD dementia	MCI or mild AD dementia	MCI or mild AD dementia
Cognitive Testing	MMSE 22–30; impaired episodic memory ≥1 SD below norms (WMS)	MMSE 22–30 (or equivalent)	MMSE 20–28	MMSE 20–30 (or equivalent)
BMI	17-35	-	-	-
Biomarkers	Amyloid positive (PET or CSF)	Amyloid positive (PET or CSF)	Amyloid PET ≥37 Centiloids; tau PET required	Amyloid positive (PET or CSF)
MRI Hemorrhage Criteria	>4 microhemorrhages (defined as ≤10 mm at greatest diameter) single macro hemorrhage >10 mm any superficial siderosis vasogenic edema contusion/encephalomalacia/aneurysms/vascular malformations/infective lesions multiple lacunar infarcts OR stroke involving a major vascular territory severe small vessel/white matter disease space-occupying lesions/brain tumors (with small meningioma/arachnoid cyst exceptions)	More than 4 microhemorrhages (defined as 10 millimeter [mm] or less at the greatest diameter); a single macro hemorrhage >10 mm at greatest diameter; an area of superficial siderosis; evidence of vasogenic edema; more than 2 lacunar infarcts or stroke involving a major vascular territory; severe subcortical hyperintensities consistent with a Fazekas score of 3 (60); evidence of amyloid beta-related angiitis (ABRA); cerebral amyloid angiopathy-related inflammation (CAA-ri); or other major intracranial pathology that may cause cognitive impairment	ARIA-E >4 cerebral microhemorrhages >1 area of superficial siderosis any macro hemorrhage severe white matter disease	Presence on baseline MRI of amyloid-related imaging abnormalities of edema/effusion, more than 4 cerebral microhemorrhages, any area of superficial siderosis, any intracerebral hemorrhage greater than 1 cm or severe white matter disease, CAA-related inflammation, territorial infarcts > 1 cm, >2 lacunar infarcts, cerebral contusion, encephalomalacia, brain aneurysms or other vascular malformations
Recent Neurologic Events	Stroke/TIA/seizure ≤12 months	Stroke/TIA/seizure ≤12 months	Stroke/TIA/seizure ≤12 months	Stroke/TIA/seizure ≤12 months
Neurological Exclusions	Any neurological condition that may be contributing to cognitive impairment above and beyond that caused by the subject’s AD	Any medical, neurologic, or psychiatric condition that may be contributing to cognitive impairment or any non-AD MCI or dementia	Significant neurological disease affecting the CNS	Non-AD neurologic condition that may be significantly contributing to cognitive or behavioral impairment
Psychiatric Exclusions	Major psychiatric illness; GDS >8	Mild psychiatric illness allowed if stable; severe excluded, including major depression	Psychiatric illness interfering with participation; history of schizophrenia or other chronic psychosis; suicide risk	Psychiatric disorder, suicidal ideation that interferes with comprehension of the requirements, potential benefit, and potential harms of treatment
MRI scanning Exclusions	Contraindication to MRI scanning	Contraindication to MRI scanning	Contraindication to MRI scanning	Contraindication to MRI scanning
Bleeding Risk	Bleeding disorder; INR >1.5; platelets <50,000	Exclude anticoagulant use/bleeding disorder (INR >1.5, platelets <50K)	Bleeding disorder; INR >1.5; anticoagulation excluded	Exclude anticoagulant use/bleeding disorder (INR >1.5, platelets <50K)
Immunologic Disease	Uncontrolled or requiring biologics; immunosuppressive therapy excluded	Any history of immunologic disease (e.g., lupus erythematosus, rheumatoid arthritis, Crohn’s disease) or systemic treatment with immunosuppressants, immunoglobulins, or monoclonal antibodies or their derivatives	Immunologic disease; immunosuppressive therapy excluded	Any history of immunologic disease (e.g., lupus erythematosus, rheumatoid arthritis, Crohn’s disease) or systemic treatment with immunosuppressants, immunoglobulins, or monoclonal antibodies or their derivatives
Cancer	Within 3 yrs (except low risk)	-	Within 5 yrs (except low risk)	Active cancer that interferes with the ability to comply with donanemab treatment
Substance Use	Alcohol/drug use disorder within 2 yrs	-	Alcohol/drug use disorder within 2 yrs	alcohol or substance use that interferes with comprehension of the requirements, potential benefit, and potential harms of treatment
Other Clinical Abnormality (medical exclusion)	Clinically important abnormality; TSH/B12 abnormalities; HIV; planned surgery; severe sensory impairment; planned surgery	Unstable medical conditions that may affect or be affected by lecanemab therapy	Serious unstable systemic illness; life expectancy <24 months; liver function >2.5× normal; drug allergy history	Unstable medical conditions.

Abbreviations: AD: Alzheimer’s disease; MCI: mild cognitive impairment; CDR: Clinical dementia scale; MMSE: Mini-Mental State Examination; WMS: Wechsler Memory Scale IV-Logical Memory II; BMI: body mass index; CNS: Central Nervous System; ARIA-E: Amyloid-related imaging abnormalities–edema; ABRA: Amyloid beta–related angiitis.

Relevant clinical data were extracted from the clinical records, including demographics, age at symptom onset, time from symptom onset to initial assessment, body mass index (BMI), scores on bedside tests of cognition, formal neuropsychological reports, biomarker results (CSF and amyloid-PET results), and stage of cognitive impairment as measured by the global Clinical Dementia Rating^® (CDR) ²⁶. Sixty-four of 184 (34.8%) patients had a global CDR score assigned at the time of the clinic visit. In other participants (120/184, 65.2%), global CDR scores were retrospectively assigned, referencing clinical history. Contemporaneous neuropsychological assessments were available for 123/184 (66.8%) participants. Participants for whom global CDR scores could not be reliably assigned were excluded (n = 3). Prior work has demonstrated excellent agreement between CDR scoring based on comprehensive medical record review and face-to-face assessment, with intraclass correlation coefficients of 0.92–0.95 for global CDR and CDR Sum-of-Boxes ²⁷.

Electronic health records were reviewed for comorbid conditions relevant to trial eligibility, including autoimmune and immunologic disorders and/or treatments, history of malignancy, uncontrolled psychiatric illness, known or suspected substance abuse within the prior two years, and cardiopulmonary conditions within the preceding year that were not stably or adequately controlled or that could affect safety or interfere with treatment (per lecanemab exclusion criteria) ¹. Cardiopulmonary review included active congestive heart failure, symptomatic coronary artery disease within 12 months, angina, significant cardiac rhythm abnormalities, or other severe cardiopulmonary conditions, as previously described ⁴. Charts were also screened for alternative neurological conditions that could contribute to cognitive impairment (CNS-related exclusion), history of seizure or stroke within 12 months of evaluation, and contraindications to MRI.

Imaging-based exclusion criteria were applied referencing radiology reports with direct review of brain images (D.S.). MRI scans were specifically assessed for clinically relevant structural lesions, cerebral microhemorrhages, areas of superficial siderosis, encephalomalacia, multiple lacunar infarcts, infarction involving a major vascular territory, severe small vessel disease or white matter disease, and intracranial tumors. Imaging abnormalities used to determine ineligibility were documented in formal radiology reports, with the exception of 2 cases in whom severe white matter changes were established on direct review. In patients with vascular malformations or other focal lesions, lesion size was measured to determine whether exclusion thresholds were met. Patients with unavailable imaging or missing gradient-recalled echo or susceptibility weighted imaging sequences were excluded.

Although the exclusion criteria were largely similar across the four eligibility frameworks, there were important differences in inclusion criteria that contributed to variation in eligibility rates (Table 1). CLARITY-AD required objective memory impairment as demonstrated by a memory CDR score ≥0.5, Wechsler Memory Scale (WMS) score (available for 96/184, 52.2% patients in the cohort), and a specified BMI range (17-35 kg/m²). Cognitive thresholds also differed between frameworks: CLARITY-AD and the lecanemab appropriate use recommendations applied stricter MMSE cutoffs (22-30), whereas the donanemab appropriate use recommendations permitted a broader MMSE range (20–30). Additionally, the clinical trials required a minimum duration of symptoms prior to enrollment (six months for CLARITY-AD, one year for TRAILBLAZER-ALZ2). Age restrictions varied across frameworks.

AD Confirmation

Included participants met established criteria for AD pathology based on CSF (129/184, 70.1%) and/or amyloid-PET (48/184, 26.1%) biomarkers, apart from four CBS-AD participants and three lvPPA participants, in whom the diagnosis of AD was established upon death (brain autopsy confirming AD neuropathology without co-pathology).

For CSF-based confirmation, biomarker profiles were extracted from clinical reports. Some (n=25) of the CSF samples were analyzed externally at Athena Diagnostics (Worcester, MA). Clinical reports from Athena Diagnostics provided individual CSF analyte levels, along with an Aβ42/t-tau index (ATI), calculated as Aβ42/(240 + 1.18 t-tau). Athena Diagnostics reports abnormal values (indicative of AD) as ATI <0.8 and p-tau >61 pg/mL ²⁸. The remaining CSF samples (n=104) were analyzed internally at Mayo Clinic Laboratories (Rochester, MN). Previously validated CSF p-tau181/Aβ42 ratio cut points were used. Roche Elecsys Generation 1 assays (March 2020–May 2023) used a positivity threshold of ≥0.023, while Generation 2 assays (from June 2023) used a threshold of ≥0.028 ^29,30.

Amyloid PET imaging was performed using Carbon-11 Pittsburgh Compound B (¹¹C-PiB) with previously described methods ³¹. Scans were considered positive if the Centiloid value was ≥25 (SUVR ≥1.52), consistent with an intermediate-to-high level of AD neuropathologic change. SUVR values were calculated by normalizing target regions to cerebellar crus gray matter ³².

Statistical analysis

Patient characteristics were summarized using descriptive statistics. Fisher's exact and Kruskal-Wallis tests were used to determine differences in demographics and clinical characteristics. Pairwise Fisher’s exact tests were used to compare eligibility across atypical phenotypes and effect sizes are presented as odds ratios. To assess whether the distribution of each ineligibility reason differed across phenotypes, the Fisher–Freeman–Halton exact test was used due to small cell counts, with two-sided p-values estimated via Monte Carlo simulation (20,000 iterations). Statistical analyses were conducted using SPSS software (version 28.0.1.1; IBM Corp., Armonk, NY). Tests were two-sided; statistical significance was defined as p < 0.05.

Data availability

The data that supports the findings of this study are available from the corresponding author, upon reasonable request from a qualified investigator. Link to MCSA/ADRC data request: https://www.mayo.edu/research/centers-programs/alzheimers-disease-research-center/research-activities/mayo-clinic-study-aging/for-researchers/data-sharing-resources.

Results

Patient characteristics

235 patients with atypical AD were screened (Figure 1). Fifty-one were excluded due to insufficient imaging (n=34), inconclusive or inconsistent AD biomarker results (n=10), or missing clinical data required for eligibility assessment (n=7), resulting in a final cohort of 184 patients. Table 2 summarizes the demographic features according to phenotype. The study cohort included 184 atypical AD patients, including 98 (53.3%) individuals with PCA, 42 (22.8%) with lvPPA, 37 with dAD (20.1%). And 7 (3.8%) with CBS-AD. Overall, 61.4% were female, with similar sex distribution across phenotypes (p = 0.372). The majority of participants self-reported race and ethnicity as White non-Hispanic (n=173, 94.0%). Race and ethnicity were self-reported distributed as follows: Asian (n = 1, 0.5%), African American (n = 2, 1.1%), Native Hawaiian (n = 1, 0.5%), other/unknown (n = 4, 2.2%), White Hispanic (n = 3, 1.6%), and White non-Hispanic (n = 173, 94.0%).

Figure 1. Flow diagram illustrating cohort selection.

Abbreviations: lvPPA = logopenic variant primary progressive aphasia; PCA = posterior cortical atrophy; CBS-AD = corticobasal syndrome due to Alzheimer’s disease; dAD = dysexecutive Alzheimer’s disease; CDR = Clinical Dementia Rating; MMSE = Mini-Mental State Examination.

Table 2. Characteristics table according to phenotype.

p. values derived from chi-square tests for categorical variables and Kruskal-Wallis tests for continuous variables. Abbreviations: lvPPA = logopenic variant primary progressive aphasia; PCA = posterior cortical atrophy; CBS-AD = corticobasal syndrome due to Alzheimer’s disease; dAD = dysexecutive Alzheimer’s disease; CDR = Clinical Dementia Rating; MCI = mild cognitive impairment; MMSE = Mini-Mental State Examination; BMI = body mass index; SD = standard deviation.

Characteristic	Full cohort	lvPPA	PCA	CBS- AD	dAD	p- value
N	184	42	98	7	37
Sex
Female, n (%)	113 (61.4)	21 (50.0)	63 (64.3)	5 (71.4)	24 (64.9)	0.372
Male, n (%)	71 (38.6)	21 (50.0)	35 (35.7)	2 (28.6)	13 (35.1)
Education, years, mean ± SD	15.7 ± 3.1	16.4 ± 2.9	15.6 ± 3.6	16.4 ± 2.1	15.2 ± 2.3	0.122
Stage of cognitive impairment at assessment
None, CDR 0, n (%)	4 (2.2)	3 (7.1)	0 (0)	1 (14.3)	0 (0)	0.005
MCI / very mild (CDR 0.5), n (%)	83 (45.1)	30 (71.4)	37 (37.8)	2 (28.6)	13 (35.1)
Mild dementia (CDR 1), n (%)	68 (37.0)	7 (16.7)	41 (41.8)	2 (28.6)	17 (45.9)
Moderate dementia (CDR 2), n (%)	27 (14.7)	2 (4.8)	18 (18.4)	1 (14.3)	6 (16.2)
Severe dementia (CDR 3), n (%)	2 (1.1)	0 (0)	2 (2.0)	0 (0)	0 (0)
Age at symptom onset, years, median (IQR)	58.7 (53.3-65.5)	63.1 (56.6-65.2)	58.3 (53.2-65.2)	56.1 (55-62.3)	55.1 (50.8-57.1)	0.039
Time from symptom onset to presentation, years, mean ± SD	2.9 (1.8-4.1)	3.2 (2-4.1)	2.9 (1.6-4.1)	3 (2.1-3.3)	2.2 (1.6-4.3)	0.634
Age at presentation, years, mean ± SD	62 (57.3-68.4)	66.9 (59.3-73)	61.2 (57.1-67.6)	60.6 (58.4-64.7)	57.3 (54.2-60.3)	<0.001
MMSE score, median (IQR)	21 (17-24)	21 (16-24.8)	21 (17-23)	27 (23.5-29)	22 (17-25)	0.036
BMI, median (IQR)	24.6 (21.7-27.6)	25.3 (22.1-27.8)	24.5 (22.1-27.8)	24.9 (22.6-27.7)	24.3 (20.7-26.8)	0.559

Stage of cognitive impairment differed across phenotypes (p=0.005), with lvPPA patients more frequently presenting at earlier stages (71.4% CDR 0.5), whereas PCA and dAD more often presented at mild-to-moderate dementia stages (PCA: 60.2% CDR ≥1; dAD: 62.1% CDR ≥1); CBS-AD showed a more variable distribution, though numbers were small. Despite these differences, time from symptom onset to diagnosis did not differ between groups (p = 0.634). Moderate dementia (CDR 2) was observed in 14.7% of the overall cohort; severe dementia (CDR 3) was rare (1.1%).

Age at assessment differed across phenotypes, with lvPPA patients oldest (median [IQR] 66.9 [59.3–73.0]) and dAD patients youngest (57.3 [54.2–60.3], p<0.001). Early-onset AD (defined as symptom onset at or before age 65) was observed in 142 of 184 patients (77.2%). MMSE scores also differed across phenotypes, with higher scores in CBS-AD (27 [23.5–29]) and lower scores in PCA (21 [17–23]) and lvPPA (21 [16–24.8], p=0.036).

Eligibility

Overall, 27/184 patients (14.7%) met CLARITY-AD eligibility criteria and 31 (16.8%) met TRAILBLAZER-ALZ2 eligibility criteria. Eligibility increased when applying appropriate use criteria, with 49 patients (26.6%) eligible for lecanemab and 59 (32.1%) for donanemab (Table 3, Figure 2).

Table 3. Eligibility Across Trial Criteria and Appropriate Use Recommendations.

Number and percentage of patients meeting eligibility criteria under each framework. Percentages are calculated within each phenotype (row denominator = N for that phenotype). Abbreviations: lvPPA = logopenic variant primary progressive aphasia; PCA = posterior cortical atrophy; CBS-AD = corticobasal syndrome due to Alzheimer’s disease; dAD = dysexecutive Alzheimer’s disease.

Phenotype	N	CLARITY- AD n (%)	TRAILBLAZ ER-ALZ2 n (%)	Lecanemab Appropriate Use Recommendations n (%)	Donanemab Appropriate Use Recommendations n (%)
Full cohort	184	27 (14.7%)	31 (16.8%)	49 (26.6%)	59 (32.1%)
PCA	98	12 (12.2%)	12 (12.2%)	20 (20.4%)	27 (27.6%)
lvPPA	42	3 (7.1%)	11 (26.2%)	13 (31.0%)	15 (35.7%)
dAD	37	12 (32.4%)	6 (16.2%)	13 (35.1%)	14 (37.8%)
CBS-AD	7	0 (0%)	2 (28.6%)	3 (42.9%)	3 (42.9%)

Figure 2. Eligibility Across Trial and Appropriate Use Frameworks According to Phenotype.

Proportion of patients meeting eligibility criteria for CLARITY-AD, TRAILBLAZER-ALZ2, and appropriate use recommendations (AUR) for lecanemab and donanemab across atypical Alzheimer’s disease phenotypes and the full cohort. Abbreviations: PCA = posterior cortical atrophy; lvPPA = logopenic variant primary progressive aphasia; dAD = dysexecutive Alzheimer’s disease; CBS-AD = corticobasal syndrome due to Alzheimer’s disease; AUR = Appropriate Use Recommendations.

Eligibility differed across atypical AD phenotypes for CLARITY-AD, with dAD patients more likely to meet trial criteria compared with lvPPA (32.4% vs 7.1%; OR 6.26, 95% CI 1.64–23.9, p = 0.008) and PCA (32.4% vs 12.2%; OR 3.43, 95% CI 1.38–8.55, p = 0.011). For TRAILBLAZER-ALZ2, lvPPA patients were more likely than PCA patients to meet eligibility criteria (26.2% vs 12.2%; OR 2.53, 95% CI 1.05–6.08, p = 0.049). No statistically significant differences in eligibility across phenotypes were observed for lecanemab or donanemab appropriate use recommendations.

Across frameworks, the most common drivers of ineligibility were cognitive thresholds, imaging-based exclusions, and symptomatic stage/severity (Figure 3). In CLARITY-AD, MMSE criteria were the largest contributor to ineligibility (59.9%), followed by imaging exclusions (22.3%) and memory/stage requirements (memory subdomain of the CDR and global CDR severity criteria; ~19% each). Notably, 94/184 (51.1%) patients were excluded from CLARITY-AD and lecanemab appropriate use criteria due to low MMSE scores; 63 of these individuals (67%) had a global CDR of 0.5 or 1, indicating early symptomatic disease.

Figure 3. Reasons for Ineligibility Across Anti-Amyloid Clinical Trial and Appropriate Use Recommendations.

Horizontal bar plots showing the distribution of exclusion criteria among patients deemed ineligible under each framework. Because multiple criteria could apply to a single individual, percentages for exclusion reasons are not mutually exclusive. Abbreviations: MMSE = Mini-Mental State Examination; WMS = Wechsler Memory Scale; BMI = body mass index.

For TRAILBLAZER-ALZ2 framework, age restrictions were the most frequent reason for ineligibility (56.1%), followed by exclusions due to MMSE criteria (50.3%), imaging abnormalities (22.6%), and disease stage (i.e., CDR >1; 18.7%). Similarly, 77/184 patients (41.8%) were excluded from TRAILBLAZER-ALZ2 based on MMSE thresholds, including 47 individuals (61%) with early symptomatic disease (i.e., global CDR 0.5–1).

Within the appropriate -use criteria, MMSE thresholds remained the predominant driver of ineligibility for lecanemab (67.4%) and donanemab appropriate use criteria (56.2%), followed by imaging exclusions (25.4–27.3%) and CDR-based severity (21.0–22.7%). Across all frameworks, comorbid conditions, including alternative neurological diagnoses (~10–12%), immunologic disease (~9–11%), and recent seizure/stroke (~8–10%), accounted for smaller proportions of exclusions, whereas anticoagulation, malignancy, psychiatric illness, and MRI contraindications were relatively infrequent contributors. There were no differences noted when comparing reasons for ineligibility across atypical phenotypes (eTables 1 and 2) but for a global CDR score of 0 - observed more frequently in the lvPPA phenotype.

Imaging-based exclusions were identified in 37/184 (20.1%) patients. These included severe small vessel disease/white matter changes or multiple lacunar infarcts (n = 17), >4 microhemorrhages or a single macro hemorrhage >10 mm (n = 10), superficial siderosis (n = 4), encephalomalacia (n = 2), cerebral amyloid angiopathy-related inflammation (CAA-ri) or other suspected inflammatory processes (n = 2), and aneurysm exceeding protocol limits or vascular malformation (n = 3). Some patients had more than one exclusionary imaging finding.

Discussion

In this retrospective eligibility assessment of a biomarker-confirmed cohort of atypical AD, approximately 70–85% of patients with atypical AD phenotypes would not have met current eligibility criteria for AAT. Most exclusions were driven by MMSE-based exclusion rather than CDR. Stated more plainly, although 84.2% of patients with atypical AD presented at an early disease stage, most would not have qualified for AAT. Importantly, this analysis was retrospective and theoretical in nature. Most patients were evaluated in routine clinical practice prior to the approval and widespread implementation of AAT and were not prospectively screened according to trial or label criteria. As such, eligibility determinations were based on available historical clinical data. Our findings therefore represent an estimate of how these patients might have aligned with eligibility frameworks if evaluated under current treatment paradigms.

Eligibility rates were broadly similar across atypical phenotypes, suggesting that current criteria systematically limit access across non-amnestic AD variants rather than disproportionately affecting a single phenotype. Although lvPPA patients were more likely than PCA patients to meet TRAILBLAZER-ALZ2 eligibility criteria, this may reflect the trial’s age restriction of 60–85 years, as a greater proportion of lvPPA than PCA patients fell within this range (71.4% vs 54.1%). Otherwise, reasons for ineligibility were largely comparable across phenotypes.

These findings highlight a potential mismatch between cognitive screening thresholds used in clinical trials and the clinical profiles of patients with atypical AD. MMSE-based exclusions were the most common reason for ineligibility across frameworks, even though global functional staging (CDR) suggested early disease in most patients. In atypical AD, domain-specific impairments (e.g., visuospatial, language, or executive dysfunction) may disproportionately lower global cognitive scores despite relatively preserved functional independence, raising concern that conventional screening tools may overestimate disease severity in these syndromes ³³. Importantly, many atypical AD phenotypes demonstrate relative hippocampal sparing in early disease stages, with correspondingly preserved episodic memory and orientation ^7,34-36. This dissociation between memory-based staging and broader cognitive dysfunction further underscores the limitations of applying uniform cognitive thresholds to non-amnestic AD variants.

When trial and appropriate use recommendations are applied to clinic populations, only 8% to 17% of patients with MCI or mild dementia are eligible for AAT^3-5. However, most prior eligibility studies have not focused specifically on biomarker-confirmed amyloid-positive patients. Instead, they have examined broader cohorts undergoing routine diagnostic evaluation or incorporated additional exclusion criteria, such as APOE genotype, limiting direct comparability with the present study ^4,5,37,38. Despite these methodological differences, restrictive eligibility criteria are widely recognized as a major barrier to access to AAT. This limitation is further compounded by comorbidities, logistical barriers, and variability in healthcare access. Experience from the Mayo Clinic Alzheimer’s Disease Treatment Clinic suggests that eligibility rates may be higher within specialized referral populations: approximately 30% of newly referred patients were deemed eligible for AAT after comprehensive multidisciplinary evaluation ³⁸. A fraction of treated patients had atypical clinical syndromes (9 of 55, 16%) or were under the age of 65 (6 of 55, 11%) at the time of therapy initiation. Reasons for ineligibility included advanced disease stage, with 11.6% of patients presenting with moderate dementia, while MRI abnormalities were identified in 21% of patients and frequently served as exclusion criteria.

Given that patients with atypical AD are more likely to present with symptom onset before 65 years-of-age, one might anticipate that comorbidities and vascular exclusions would play a lesser role in limiting eligibility. However, vascular-related exclusions remained a meaningful contributor to ineligibility in our cohort, with 22.3-27.3% of patients excluded due to vascular imaging findings. Although this proportion is somewhat lower than the 31.7–37.5% reported in cohorts of typical AD (notably among patients who already met trial inclusion criteria)⁴, it nonetheless highlights the substantial impact of vascular pathology even in younger, atypical populations. In particular, cerebral amyloid angiopathy may contribute to this burden, as prior studies suggest that atypical AD is associated with increased microbleed burden ^39,40.

Patients with early-onset and atypical AD consistently demonstrate a higher burden of tau pathology, particularly in neocortical and hub regions, compared to late-onset or typical AD ^7,8,41-44 This increased tau burden has been associated with more rapid clinical decline and greater executive dysfunction ⁴⁵. In post hoc analyses of trial data, participants with lower tau burden derived greater benefit from donanemab, whereas those with higher tau burden demonstrated less clinical improvement ^46,47. Accordingly, it has been proposed that AAT, which target amyloid-β, may be less effective in these patients, whose disease trajectory may be more strongly driven by tau-mediated neurodegeneration than by amyloid burden ⁴⁸. This hypothesis remains speculative and has not been directly tested, in large part because patients with atypical phenotypes and early-onset AD continue to be underrepresented in clinical trials ^1,2 and in ADTC populations ⁴⁹. Thus, even if phenotype-sensitive cognitive and severity thresholds were developed, the efficacy and safety of AAT in atypical AD would still need to be established carefully in prospective studies before broader clinical adoption could be justified. Although recent trials, including donanemab studies⁵⁰ incorporating tau PET stratification, have broadened inclusion, these studies were not specifically powered or designed to assess efficacy and safety across distinct atypical AD phenotypes. As a result, phenotype-specific treatment effects and risk-benefit profiles remain incompletely characterized.

There are limitations to our study. Global CDR was retrospectively assigned in a subset of patients, which may introduce classification bias. APOE genotype was not incorporated into eligibility assessments and could further reduce eligibility under current prescribing practices. The CBS-AD subgroup was small, limiting statistical power for phenotype-specific comparisons. In addition, MRI data were acquired across different scanners and protocols, which may introduce variability in imaging measures. Finally, this was a retrospective, single-center analysis within a tertiary referral population, which may limit generalizability.

Recent advances in the field, including the rapid integration of blood-based biomarkers, advanced imaging modalities, and AAT into routine clinical practice, have transformed care for many patients with typical amnestic presentations of AD. However, these developments have also exposed important gaps for individuals with atypical presentations. Our findings underscore the need to critically re-evaluate the validity and clinical relevance of cognitive and severity thresholds when applied to non-amnestic AD variants. In clinical practice, treatment eligibility is more nuanced than trial screening alone and may incorporate factors beyond bedside cognitive scores, including functional status, age, comorbidities, and individualized risk-benefit considerations; as such, some patients with atypical AD who would not have met trial criteria may still be considered for treatment. Future studies should evaluate atypical AD both in dedicated trials and as prespecified subgroups within larger therapeutic studies. Refining eligibility frameworks to better reflect the clinical realities of atypical AD will be essential not only to ensure equitable access to treatment, but also to appropriately assess therapeutic benefit in these underrepresented populations.