Clinical and MRI features associated with amyloid positivity in secondary screening for anti–amyloid‐β therapy

Abstract

INTRODUCTION

In Japan, confirmatory biomarker testing for Alzheimer's disease (AD) is insurance‐covered only when evaluating eligibility for anti–amyloid‐β therapies. This study aimed to identify clinical and magnetic resonance imaging (MRI) features associated with amyloid positivity in a real‐world setting.

METHODS

We reviewed patients attending our secondary screening clinic between 2023 and 2024. After clinical discussions, interested patients underwent 3T MRI and either amyloid positron emission tomography (PET) or cerebrospinal fluid biomarker testing.

RESULTS

Of the 200 patients, 147 proceeded to evaluation and 138 had biomarker results. The amyloid‐positive group (n = 107) was younger and had lower Mini‐Mental State Examination (MMSE) scores, despite similar symptom duration; younger age largely reflected early‐onset cases. MRI indices overlapped between the groups, except for a higher medial‐temporal‐whole‐brain atrophy ratio.

DISCUSSION

Younger age and lower MMSE scores were associated with amyloid positivity. Caution regarding referral bias for early‐onset AD in real‐world settings is warranted.

Highlights

• Research in a real‐world secondary screening for anti–amyloid‐β therapy.

• Younger age and lower Mini‐Mental State Examination was associated with amyloid positivity.

• Medial temporal lobe atrophy was common among amyloid‐negative patients.

• Medial‐temporal–whole‐brain atrophy ratio was the only difference in magnetic resonance imaging indices.

• Simple clinical features may guide Alzheimer's disease prescreening before biomarker testing.

1. INTRODUCTION

Although Alzheimer's disease (AD) is clinically well known for gradual cognitive decline leading to dementia, it is defined neuropathologically by the presence of senile plaques and neurofibrillary tangles in the brain, ¹ which are aggregates of amyloid beta (Aβ) ² and tau, ³ , ⁴ respectively. Approximately 30% of patients clinically diagnosed with “AD dementia” may not have underlying AD pathology. ⁵ Even in clinical practice, the field is moving toward diagnoses based on biomarkers that reflect brain pathology, ⁶ particularly now that anti‐Aβ antibody therapies (AATs) are available. Because blood‐based biomarkers are approved only in some countries, cerebrospinal fluid (CSF) biomarkers or amyloid positron emission tomography (PET) are mainly used as confirmatory tests. However, because PET is expensive and CSF collection is burdensome, these confirmatory tests are often conducted in a limited number of patients, such as those interested in AATs. In Japan, amyloid PET, CSF Aβ42/40, and CSF phosphorylated tau 181 (p‐tau181)/Aβ42 are covered by insurance only for the evaluation of AAT eligibility.

Although AATs have demonstrated efficacy in clinical trials, ⁷ , ⁸ not all patients choose to receive these therapies. Shared decision‐making is therefore extremely important. ⁹ , ¹⁰ However, such discussions may be complicated by the fact that amyloid positivity has not yet been confirmed at the time of consultation, and, thus, the diagnosis of AD has not yet been established. Even an approximate estimate of likelihood can be informative for shared decision‐making. For example, if a patient understands that the likelihood of AD is extremely high, the patient may be more inclined to pursue confirmatory biomarker testing and subsequent treatment. Conversely, if a patient is relatively hesitant and understands that the likelihood of AD is low, the patient may choose not to proceed with further testing or may even opt for prescreening when blood‐based biomarkers become available, since the negative predictive value of the test becomes high when the pretest probability is low. Therefore, there is an important need to estimate the approximate likelihood of amyloid positivity before performing these tests. Although previous investigations have attempted to infer the amyloid status based on clinical characteristics, cognitive performance, and neuroimaging findings, ¹¹ , ¹² to the best of our knowledge, no study has characterized the clinical profiles of patients evaluated for AAT eligibility in a real‐world setting.

Patients with neurodegeneration (N+) suspected of having symptomatic AD but showing negative AD biomarker results are often described as suspected non‐AD pathophysiology. ¹³ Such conditions—including primary age‐related tauopathy (PART), ¹⁴ limbic‐predominant age‐related TAR DNA‐binding protein of 43 kDa (TDP‐43) encephalopathy, ¹⁵ and argyrophilic grain disease (AGD) ¹⁶ —are characterized by medial temporal lobe (MTL)‐predominant neurodegeneration and have been reported to preferentially affect older individuals while being associated with relatively preserved cognitive function. ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ However, the clinical and magnetic resonance imaging (MRI) characteristics of amyloid‐negative non‐AD pathologies in the real‐world evaluation for AATs, including the Voxel‐based Specific Regional Analysis system for Alzheimer's Disease (VSRAD), ¹⁸ remain unknown.

Therefore, in this study, we evaluated the clinical and MRI characteristics associated with amyloid positivity and negativity at a secondary screening clinic for AATs.

2. METHODS

2.1. Study population and design

This study was a single‐center, retrospective, cross‐sectional study designed to examine the association of amyloid status with clinical and MRI features at a secondary screening clinic for AATs. We reviewed all patients who visited our AD disease‐modifying therapy (DMT) clinic, a secondary screening specialty clinic within our institution, between December 2023 and October 2024. The situation of AATs in Japan, details of our two‐step screening process, and the reasons for screening failures at each step have been reported previously. ¹⁹ We selected this study period because it preceded the modification of our initial screening requirement in November 2024, which expanded eligibility to include patients with Mini‐Mental State Examination (MMSE) scores of 20–21 following the launch of donanemab on November 26, 2024. Briefly, the requirement for referral included a diagnosis of mild cognitive impairment (MCI) or mild dementia, MMSE score ≥22 (as mandated by the Japanese Optimal Use Guideline for lecanemab), no evidence of disease other than AD primarily causing cognitive impairment, and the ability to undergo MRI. ¹⁹ Patients interested in treatment subsequently underwent 3T MRI, Clinical Dementia Rating (CDR; also mandated by the Japanese Optimal Use Guideline), and amyloid PET or CSF biomarker testing.

Symptom duration was determined by reviewing patients’ medical records and was defined as the number of years from the onset of symptoms, such as memory impairment, at the initial visit to the AD DMT clinic. If the interval from symptom onset to the clinic visit was less than 1 year, the duration was calculated by dividing the number of months by 12. “MMSE point loss/duration” was calculated as (30 – MMSE score) divided by symptom duration.

2.2. Ethical approval

This study was performed in accordance with the tenets of the Declaration of Helsinki and was approved by the Institutional Review Board of the Tokyo Metropolitan Institute for Geriatrics and Gerontology (TMIG) (R23‐117).

2.3. MRI protocol and volumetric analysis

All MRI scans were obtained using the same 3T MRI scanner (Ingenia Elition 3.0T; Koninklijke Philips NV, Amsterdam, Netherlands). Volumetric analyses based on three‐dimensional T1‐weighted images were conducted using VSRAD Advance 2 software based on SPM8 plus DARTEL, ¹⁸ a free software package widely implemented in routine clinical practice in Japan. In this system, the mean of positive of z‐score values within the MTL volume of interest (VOI) is calculated as the “VOI severity.” In addition, the “VOI extent” is defined as the percentage of voxels within the VOI that exceed the threshold z‐score, whereas the “GM extent” represents the percentage of all gray matter (GM) voxels exceeding the threshold. The “VOI/GM ratio” is then derived as the proportion of VOI extent to GM extent, reflecting the relative selectivity of atrophy in the MTL region. The “asymmetry index” was defined as the absolute difference between the left and right VOI severity values divided by the VOI severity. The asymmetry index was devised for VSRAD analyses with reference to previously published methods. ²⁰

RESEARCH IN CONTEXT

1. Systematic review: We searched PubMed for studies on the prediction of amyloid positivity using baseline clinical or magnetic resonance imaging (MRI) data and real‐world screening for anti–amyloid‐β (anti‐Aβ) antibody therapies. Although several institutions reported real‐world screening experiences for anti‐Aβ antibody treatments, we did not identify any study that specifically focused on determining the factors associated with amyloid positivity in the context of anti‐Aβ therapies in routine clinical practice, at the time of searching.

2. Interpretation: In patients who passed primary screening (Mini‐Mental State Examination [MMSE] ≥22) and underwent secondary evaluation for anti‐Aβ therapies, younger age and lower cognitive scores—despite similar symptom duration—were associated with amyloid positivity. Medial temporal atrophy was common among amyloid‐negative individuals, consistent with suspected non‐Alzheimer's disease pathophysiology, and was not specific for Alzheimer's disease.

3. Future directions: Larger multicenter studies should validate these findings and refine prescreening strategies to better identify appropriate candidates for anti‐Aβ therapies.

2.4. Amyloid PET

All amyloid PET tests conducted during this period used ¹⁸F‐flutemetamol (Vizamyl; Nihon Medi‐Physics, Tokyo, Japan). Static emission data were acquired 90–110 min after intravenous bolus injections, with an administered dose of ≈185 MBq. Visual reads were performed by an expert (K. Ishii). For patients who had undergone prior amyloid PET using other approved methods, the images were re‐evaluated and confirmed to be visually positive by the same expert reader.

2.5. CSF AD biomarker measurement

CSF was obtained via a standard lumbar puncture. During the study period in Japan, the Aβ42/Aβ40 ratio determined using the LUMIPULSE system (FUJIREBIO INC., Tokyo, Japan) was the only approved CSF assay to confirm eligibility for lecanemab. CSF concentrations of Aβ42 and Aβ40 were measured using the LUMIPULSE system and its assays. ²¹ , ²² The approved predetermined single cutoff value in Japan is 0.67. ²²

2.6. Statistical analysis

All statistical analyses were performed using R software (version 4.5.0; R Foundation for Statistical Computing, Vienna, Austria). Continuous variables were summarized as means ± standard deviations (SDs) or as medians and interquartile ranges (IQRs) (25th–75th percentile) and were compared between groups using the Wilcoxon rank‐sum test. Categorical variables were expressed as absolute numbers and percentages (%) and compared using Pearson's chi‐square test or Fisher's exact test, as appropriate. For analyses restricted to participants 65 years of age or older, the same descriptive statistics and group comparisons were performed to minimize heterogeneity arising from differences between early‐onset AD (EOAD) and late‐onset AD. ²³ , ²⁴ To account for multiple comparisons across continuous variables in this subgroup analysis, p‐values were adjusted using the Holm method.

To assess whether the amyloid status modifies the association between neuroimaging measures and cognition in participants 65 years of age or older, we used multiple linear regression with MMSE as the dependent variable. The predictors included amyloid status, mean‐centered VOI extent, and their interaction, with adjustments for GM extent, age, and symptom duration. The model is specified as follows:

$$\begin{array}{ccc}\hfill MMSE& =& {\beta }_0+{\beta }_1\times Amyloid+{\beta }_2\times VOIexten{t}_c+{\beta }_3\times Amyloid\hfill \\ & & \times VOIexten{t}_c+{\beta }_4\times GMextent+{\beta }_5\times Age\hfill \\ & & +{\beta }_6\times Symptomduration+\epsilon \hfill \end{array}$$

Here, VOI extent _c denotes the mean‐centered VOI extent, Amyloid is coded as 1 for A+ and 0 for A−, and ε is the error term.

To quantify the cognitive performance relative to that expected from demographic and neuroimaging features, we applied a residual‐based approach. The expected MMSE scores were estimated using a linear regression model, with age, symptom duration, VOI extent, and GM extent as predictors. Age was modeled using natural spline functions to account for potential nonlinear effects. The residual MMSE scores were calculated as the difference between observed and model‐predicted MMSE values. These residuals represent deviations in cognitive performance from those expected based on demographic and neuroimaging variables. Subsequently, the association between amyloid positivity and residual MMSE scores was examined using a logistic regression model, with amyloid positivity as the dependent variable. In this model, age was again adjusted using natural spline functions. Odds ratios (ORs) and 95% confidence intervals (CIs) were estimated using the profile likelihood method. The models are specified as follows:

$$\begin{array}{ccc}\hfill E\left[MMSE\right]& =& {\beta }_0+f\left(Age\right)+{\beta }_1\times Symptomduration\hfill \\ & & +{\beta }_2\times VOIextent+{\beta }_3\times GMextent+\epsilon \hfill \end{array}$$

$$logit\left(P\left(Amyloid+\right)\right)=\alpha 0+g\left(Age\right)+\alpha 1\times ResidualMMSE.$$

Here, f(·) and g(·) denote natural spline functions, and Residual MMSE = observed MMSE − E[MMSE].

3. RESULTS

3.1. Screening process

During the study period, 200 patients visited our DMT clinic. Among them, five patients showed no objective evidence of cognitive decline in the CDR. Of the remaining 195 patients, amyloid biomarker results were available for 138, who were included in the present analysis. Details of the screening procedures have been described previously. ¹⁹ Of these 138 patients, 85 (62%) underwent amyloid PET, whereas the remainder underwent CSF analysis. Amyloid positivity was confirmed in 76% of patients assessed using PET and in 79% of those assessed using the CSF. Overall, 107 patients (78%) were classified as amyloid‐positive (A+) and 31 (22%) as amyloid‐negative (A−), all of whom had MCI or mild dementia. One A− patient had a missing CDR evaluation, and another lacked MRI results. Both MRI and CDR evaluations were unavailable in two A+ and two A− cases.

3.2. Baseline characteristics and additional test results in the A+ and A− groups

Table 1 presents the baseline demographic and clinical characteristics of the overall study cohort. The A+ group was slightly but significantly younger than the A−group (74 ± 8.1 years vs 78 ± 8.6 years). The age distribution of patients is shown in Figure S1. The proportion of women and that of patients with MCI were comparable between the groups.

TABLE 1.

Baseline characteristics.

	A+ (n = 107)	A− (n = 31)	p‐value	Cohen's d [95% CI]
Baseline
Age, year	74 ± 8.1	78 ± 8.6	0.015 *	−0.42 [−0.82, −0.01]
Female	75 (70.1%)	22 (71.0%)	>0.9
MCI	64 (59.8%)	22 (71.0%)	0.26
MMSE	25 ± 2.0	26 ± 2.2	0.002 *	−0.67 [−1.09, −0.26]
Delayed recall	1.0 [0.0, 2.0]	1.0 [1.0, 2.0]	0.077	−0.37 [−0.78, 0.03]
Symptom duration, year	2.0 [1.0, 3.0]	2.0 [1.0, 4.0]	0.301	−0.18 [−0.59, 0.22]
MMSE point loss/duration	2.7 [1.6, 5.0]	1.5 [0.8, 3.5]	0.011 *	0.39 [−0.01, 0.80]
Additional test results
CDR	n = 105	n = 28
Global	0.5 [0.5, 0.5]	0.5 [0.5, 0.5]	0.172	−0.29 [−0.71, 0.13]
0.5	93 (88.6%)	22 (78.6%)
1.0	12 (11.4%)	6 (21.4%)
Memory	0.8 [0.5, 1.0]	1.0 [0.5, 1.0]	0.326	0.15 [−0.27, 0.57]
Sum of Boxes	3.0 [2.0, 3.5]	2.8 [1.5, 4.0]	0.533	0.13 [−0.29, 0.55]
MRI	n = 105	n = 28
VOI severity	1.6 [1.1, 2.2]	1.4 [0.8, 2.3]	0.411	0.05 [−0.37, 0.47]
VOI extent	29.7 [3.7, 51.2]	12.0 [0.4, 46.3]	0.268	0.16 [−0.26, 0.58]
VOI/GM ratio	6.9 [1.1, 12.8]	1.8 [0.1, 9.7]	0.036 *	0.43 [0.00, 0.85]
GM extent	3.8 [2.8, 5.2]	3.7 [2.7, 6.8]	0.508	−0.23 [−0.65, 0.20]
Asymmetry index	0.4 [0.2, 0.6]	0.3 [0.2, 0.6]	0.871	−0.07 [−0.49, 0.35]

Note: Data are presented as mean ± SD, median [IQR], or number (%), as appropriate.

Effect sizes are reported as Cohen's d with 95% confidence intervals, where negative values indicate lower values in the A+ group compared with the A− group.

Abbreviations: A−, amyloid negative; A+, amyloid positive; CDR, Clinical Dementia Rating; CI, confidence interval; GM, gray matter; IQR, interquartile range; MCI, mild cognitive impairment; MMSE, Mini‐Mental State Examination; MRI, magnetic resonance imaging; SD, standard deviation; VOI, volume of interest.

^* p < 0.05 as individual test.

The A+ group had significantly lower total MMSE scores (25 ± 2.0 vs 26 ± 2.2). Furthermore, MMSE scores were not correlated with symptom duration (Figure S2). Delayed recall scores on the MMSE tended to be lower in the A+ group. Symptom duration was similar and tended to be shorter in the A+ group (2.0 [1.0–3.0] vs 2.0 [1.0–4.0]). Consistent with these findings, the MMSE point loss per year was significantly greater in the A+ group (median [IQR]: 2.7 [1.6–5.0] vs 1.5 [0.8–3.5]) (Figure 1). CDR scores, including the sum of boxes and global ratings, did not differ significantly between the A+ and A− groups.

FIGURE 1.

Age, cognitive and structural indices by amyloid status. Box‐and‐jitter plots of age (A), MMSE (B), MMSE point loss/duration (C), and VOI/GM ratio (D) in A+ and A− participants. A+, amyloid positive; A−, amyloid negative; GM, gray matter; MMSE, Mini‐Mental State Examination; VOI, volume of interest.

Regarding MRI‐based atrophy measures obtained using VSRAD, no significant differences were observed between the A+ and A− groups in most indices, including VOI severity, VOI extent, and GM extent (Table 1). The differences remained insignificant even in patients with mild dementia (Table S1, Figure S3). VOI extent was correlated with GM extent; however, the magnitude of the association was modest (Figure S4). The VOI/GM ratio was significantly higher in the A+ group (6.9 [1.1–12.8] vs 1.8 [0.1–9.7]) (Figure 1). The right and left VOI severities are shown in Figure S5, and the asymmetry index did not differ between the groups.

To minimize the potential influence of EOAD, given the bimodal distribution in the A+ group (Figure S1), we conducted a sensitivity analysis restricted to participants 65 years of age or older (Table S2, Figure S6). The age difference between the A+ and A− groups was small and did not reach statistical significance (77 ± 5.2 years vs 79 ± 6.0 years). However, other variables demonstrated trends similar to those observed in the full cohort, including lower MMSE scores despite similar symptom durations and higher VOI/GM ratios in the A+ group. After Holm's correction for multiple comparisons, a statistically significant group difference persisted only for the MMSE scores (adjusted p = 0.011).

3.3. Association between amyloid status and MMSE scores

To assess whether the association between amyloid status and MMSE score was direct, we fitted a multiple linear regression model with the MMSE score as the dependent variable. VOI and GM extents, which are components used to calculate the VOI/GM ratio associated with A+, were included as independent variables. Their correlation was modest (r = 0.32). Even after adjusting for mean‐centered VOI extent, GM extent, age, and symptom duration, MMSE scores remained significantly lower in the A+ group than in the A− group (β = −1.69; 95% CI: −2.55 to −0.83, p = 0.00021). Other variables were not significantly associated with the amyloid status. The overall model was statistically significant (F (6, 109) = 3.90, p = 0.00145) but explained a modest proportion of the variance in MMSE (R ² = 0.18; adjusted R ² = 0.13). In simple‐slope analyses, VOI extent was not associated with MMSE in the A− group (slope = 0.003; 95% CI: −0.024 to 0.030), whereas a modest negative association was observed in the A+ group (slope = −0.018; 95% CI: −0.034 to −0.0017) (Figure 2). The between‐group differences in slopes did not reach statistical significance.

FIGURE 2.

Adjusted association between VOI extent and MMSE by amyloid status. Scatter plots depict the association between VOI extent and MMSE in participants ≥65 of years, stratified by amyloid status (A+ vs A−). Solid lines represent model‐based predicted MMSE values adjusted for age, global GM extent, and symptom duration. Shaded areas indicate the 95% confidence intervals of the predicted values. A+, amyloid positive; A−, amyloid negative; GM, gray matter; MMSE, Mini‐Mental State Examination; VOI, volume of interest.

As a secondary analysis, we fitted a linear regression model with MMSE as the dependent variable and age, symptom duration, GM extent, and VOI extent as predictors, and defined residual MMSE as the difference between observed and model‐predicted MMSE. This model showed limited explanatory power (adjusted R ² = 0.013). Higher MMSE residuals were associated with a lower likelihood of amyloid positivity (age‐adjusted OR: 0.66, 95% CI: 0.51–0.83; p = 0.0008; Figure S7).

4. DISCUSSION

In this study conducted at our DMT clinic, a real‐world secondary screening clinic for AATs, the A+ group was younger and exhibited lower MMSE scores despite having a similar symptom duration compared with the A− group. Although the VOI/GM ratio was higher in the A+ group, other VSRAD measures did not differ significantly between the two groups.

In our real‐world DMT clinic cohort, the A+ group was significantly younger than the A− group. However, when the analysis was restricted to participants 65 years of age or older to account for the potential influence of EOAD, the age difference between the groups was no longer statistically significant. This finding may be partly attributable to the relatively high proportion of EOAD in our cohort (15%, 16 of 107 participants), which exceeds the 5.5% reported in a previous meta‐analysis ²⁵ and likely reflects the referral characteristics of a DMT clinic. In a similar clinical setting such as in large referral dementia centers, our results suggest that early‐onset cases may have high likelihood of having AD. Although this pattern may in part reflect referral bias, this may represent a counterintuitive finding, given that such patients may at times be perceived as having a low pretest probability of neurodegenerative disease in routine clinical practice.

To reduce the potential effect of EOAD, we conducted a subgroup analysis limited to participants 65 years of age or older. In this subgroup, the A+ group continued to exhibit lower MMSE scores despite having similar symptom durations. Of interest, symptom duration did not differ significantly according to amyloid status. Possible explanations include underlying heterogeneity including co‐pathologies in AD, or clinical visit at relatively preserved cognitive function in the A− group. Consequently, the annualized MMSE decline was greater in the A+ group. This pattern is directionally consistent with that reported in previous studies indicating that cognitive decline is slower in A− patients than in A+ patients. ²⁶ , ²⁷ Notably, the MMSE score remained the only variable that showed a significant group difference after Holm correction for multiple comparisons, but the distributions of MMSE scores largely overlapped between the groups. Considering prior reports suggesting that anosognosia is a predictor of amyloid positivity, ²⁸ , ²⁹ patients in the A+ group may seek medical attention at a stage when cognitive function has already declined to a greater extent than in A− individuals.

Although the A+ group showed significantly lower MMSE scores than the A− group, CDR‐SB (sum of boxes) did not differ significantly between the two groups. Unfortunately, our study was underpowered to look into the details of the subdomains. We only included the memory scores and only MMSE (delayed recall) showed a trend toward a lower score, which may suggest greater memory dysfunction consist with AD ³⁰ and a difference in the dynamic range of the memory score in this early stage.

We also examined structural MRI indices in the two groups. VSRAD is an imaging tool designed to evaluate MTL atrophy, focusing on a VOI that includes the entorhinal cortex, hippocampus, and adjacent medial temporal structures, and has been reported to aid in the differentiation of clinically diagnosed AD dementia from normal aging. ¹⁸ , ³¹ However, in the present study, measures directly reflecting MTL atrophy, such as VOI severity and VOI extent, did not differ between the A+ and A− groups. Even when the analysis was restricted to participants with dementia, no significant differences were observed between groups (Table S1). This lack of group differences may be attributable to the fact that our comparison involved older individuals with cognitive impairment rather than cognitively normal elderly controls.

By contrast, the VOI/GM ratio showed a significant difference, consistent with a previous report. ³² Prior studies have also demonstrated that A+ individuals with MCI exhibit more rapid temporal cortical atrophy than A− individuals, ³³ suggesting that selective involvement of the temporal lobe may be more pronounced in the early clinical stages of AD. Nevertheless, the distributions of the VOI/GM ratio largely overlapped between the groups, indicating a limited discriminatory power for amyloid status. In this study, we focused on MRI indices reflecting MTL atrophy provided by the VSRAD software, although cortical AD‐signature VOIs may also be informative. ³⁴ Our findings suggest that assessment of MTL atrophy alone is insufficient to discriminate amyloid status.

Even after adjustment for multiple demographic, clinical, and neuroimaging covariates, MMSE scores remained significantly lower in the A+ group than in the A− group. This finding is consistent with previously reported findings, suggesting a direct or indirect cognitive impact of amyloid pathology beyond structural brain changes. ³⁵ , ³⁶ By contrast, no statistically significant difference was observed in the association between MMSE score and VOI extent according to amyloid status. Nevertheless, as illustrated in Figure 2, the relationship between VOI extent and MMSE score appeared to be stronger in the A+ group, which is directionally consistent with that reported in previous studies indicating greater cognitive decline in A+N+ individuals. ³⁷ , ³⁸

Taken together, these findings indicate that the A− group exhibited MTL atrophy comparable to that of the A+ group, tended to be older, and showed relatively preserved cognitive function. This profile is consistent with amyloid‐negative non‐AD pathologies, including PART and AGD, which are typically observed in older individuals and are characterized by MTL‐predominant neurodegeneration with relative preservation of cognition. ¹³ , ³⁹ In the secondary conceptual analysis, the model power was limited, but the significant difference in residual MMSE scores suggests that individuals whose cognitive performance is relatively preserved despite advanced age and brain atrophy are more likely to be A−. This finding is also consistent with previous evidence demonstrating that amyloid status is associated with cognitive performance. ⁴⁰ , ⁴¹ However, the finding should be interpreted with caution because it remains unclear whether the observed effect reflects amyloid‐related processes, tau‐related processes, or both.

As for the asymmetry in MTL atrophy, no significant difference was observed between the A+ and A− groups in the asymmetry index. Among amyloid‐negative non‐AD pathologies, AGD has been reported to exhibit asymmetric atrophy of the MTL, particularly involving the ambient gyrus. ¹⁶ , ⁴² , ⁴³ , ⁴⁴ However, such asymmetry was not stronger in the A− group in the present study. Several explanations may account for this finding. First, patients with AGD often have concomitant AD and asymmetrical medial temporal atrophy can also be observed in A+ individuals. ²⁰ Second, the asymmetry index derived from VSRAD may have limited sensitivity for detecting focal atrophy, predominantly affecting the ambient gyrus, because the predefined VOI is relatively broad. Future studies should incorporate more detailed analyses of MTL substructures and focus on regions that may be affected early, such as the ambient gyrus, to better capture early pathological atrophy.

This study has several limitations. First, the study was conducted in a single center in Japan, and the number of A− participants (n = 31) was considerably smaller than that of A+ participants (n = 107). Larger multicenter studies are required to validate our findings across different clinical settings. Second, amyloid biomarker results were available in only ≈70% of the patients visiting our DMT clinic, and the high rate of amyloid positivity may suggest selection bias. Therefore, the findings of this study may not be representative of earlier screening process and could be specific to specialty clinics evaluating for AATs. Third, in real‐world clinical practice, amyloid status is determined using either the CSF or PET, and some patients may show discordant results between CSF and PET analyses. ⁴⁵ Fourth, educational attainment was not included, and symptom duration was based on a retrospective chart review. Therefore, future studies using structured questionnaires are necessary. Incorporating education may improve the interpretability of residual‐based approaches, as education has been reported to contribute to cognitive resilience and cognitive test scores itself. ⁴⁶ , ⁴⁷ Fifth, this present study using VSRAD did not account for previously reported sex differences in the brain volume. ⁴⁸ Volumetric analyses adjusted for sex should be important in future studies. Finally, apolipoprotein E (APOE) genotype was not included in the study. Previous research studies have suggested that APOE genotype may be useful in predicting amyloid status. ¹¹ , ¹² However, as this study was intended to reflect prescreening in a real‐world clinical setting for AAT eligibility, APOE genotype was not incorporated into the present analysis.

In conclusion, among patients referred to our DMT clinic, the A+ group was younger and exhibited lower MMSE scores. Caution regarding referral bias for EOAD in real‐world settings is warranted. MRI indices based on VSRAD were largely similar between the groups, with the only difference observed in the VOI/GM ratio. Clinical and MRI features provide limited discriminative value for predicting amyloid status in this study, indicating limited ability of routine clinical and structural MRI assessment to determine amyloid status. However, those with preserved MMSE scores relative to age, symptom duration and MRI structural measures may have higher likelihood of being A−. Further studies are needed to better understand the clinical and MRI characteristics associated with A+ or A− in real‐world patients evaluated for AAT eligibility.

CONFLICT OF INTEREST STATEMENT

M.K. received lecture fees from Eisai, Eli Lilly, Sysmex, and John Wiley & Sons; patent assignment fees from FUJIREBIO; and research support from Nihon Medi‐Physics. T.B. received lecture fees from Eisai. A.M.T. received honoraria for lectures from Eisai and Eli Lilly. K.I. received research grant from Nihon Medi‐Physics; advisory fees, honoraria for lectures and research grants from Eli Lilly, Nihon Medi‐Physics, and PDR Pharma. R.I. received advisory fees from Eisai and Eli Lilly, and lecture fees from Eisai, Eli Lilly, PDR Pharma, Nihon Medi‐Physics, and Sysmex. A.I. has received advisory fees from Eisai and Eli Lilly; stock ownership/capital gains from Eisai; lecture fees from Eisai, Eli Lilly, John Wiley & Sons, FUJIREBIO, HU Frontier, and Sysmex; and grants for commissioned/joint research from Eisai, Eli Lilly, FUJIREBIO, and Sysmex. The other authors (K.Y., Y.A., K.S., R.S., K.H., F.S., and K.F.) have nothing to disclose related to the content of this manuscript. Author disclosures are available in the supporting information.

CONSENT STATEMENT

This retrospective study was approved by the Institutional Review Board of the Tokyo Metropolitan Institute for Geriatrics and Gerontology (TMIG) (R23‐117). Individual informed consent was waived, as the study employed an opt‐out approach in accordance with institutional ethical guidelines.

Supporting information

Supporting Information: dad270374‐sup‐0001‐SuppMat.docx

Supporting Information: dad270374‐sup‐0002‐SuppMat.pdf

DATA AVAILABILITY STATEMENT

Data supporting the findings of this study are available from the corresponding author upon reasonable request.