Associations of mild behavioral impairment and white matter hyperintensity patterns

Abstract

INTRODUCTION

It is unclear whether cerebrovascular and white matter changes contribute to mild behavioral impairment (MBI). White matter hyperintensities (WMHs) represent a prominent pathology for research due to association with cognitive decline. We investigated associations between MBI measures and WMH volume and spatial distribution.

METHODS

We analyzed 1273 individuals with normal cognition or mild cognitive impairment. General linear models and voxel‐wise analyses were employed to examine WMHs in terms of volume and spatial distribution.

RESULTS

Greater WMH volume was associated with higher MBI total score and MBI status in participants with normal cognition and mild cognitive impairment. Voxel‐wise analysis revealed a significant cluster in the left inferior longitudinal fasciculus associated with higher MBI total score.

DISCUSSION

Our findings suggest a contribution of WMH volume and spatial distribution to persistent later‐life behavioral symptoms. WMHs, especially in the left inferior longitudinal fasciculus, may serve as a neuroimaging marker of behavioral symptoms.

Highlights

• MBI score and status relate to greater WMH volume in dementia‐free individuals.

• WMHs in the left inferior fasciculus could be a potential biomarker of behavioral changes.

• Total MBI score and status were more informative than individual MBI domain scores.

• Findings support a potential contribution of white matter damage to MBI symptoms.

1. BACKGROUND

Mild behavioral impairment (MBI) was defined by the International Society to Advance Alzheimer's Research and Treatment of the Alzheimer's Association criteria as the emergence and persistence of neuropsychiatric symptoms in older adults without dementia. ¹ , ² These changes in behavior or personality are now being recognized as early indicators of dementia. ³ , ⁴ In addition, MBI is associated with an increased risk of cognitive decline and clinical conversion from normal cognition (NC) to mild cognitive impairment (MCI) and from MCI to Alzheimer's disease. ⁵ , ⁶ , ⁷ , ⁸ White matter hyperintensities (WMHs) are a common neuroimaging finding in older individuals characterized by myelin pallor, gliosis, and loss of both myelin and axons and may reflect heterogeneous processes, including cerebral small vessel disease. ⁹ , ¹⁰ Cerebral small vessel disease and WMH have been associated with the progression of neuropsychiatric symptoms, ¹¹ , ¹² cognitive decline, and dementia incidence, ¹³ , ¹⁴ suggesting that they may contribute to the development of MBI.

Most prior biomarker work in MBI focused on neurodegenerative Alzheimer's disease‐related processes, whereas the potential contribution of WMHs and white matter damage remains less well understood. ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² , ²³ Higher WMH volume has been associated with MBI status in individuals with MCI. ²⁴ In another study, it was suggested that WMHs may partially mediate MBI scores by reducing the cortical thickness in NC individuals. ²⁵ However, these studies analyzed the total WMH volume rather than their spatial distribution, and none compared associations across both NC and MCI groups within the same cohort.

This study aimed to advance the understanding of the association between MBI measures (MBI status, MBI total score, and MBI domain‐level scores) and WMHs in a large cohort of older individuals without dementia, including NC and MCI participants. We hypothesized that higher MBI total score and MBI status would be associated with greater total WMH volume in MCI as well as NC participants. We further hypothesized that WMH spatial distribution would be associated with MBI measures beyond total lesion volume, reflecting the potential involvement of specific white matter pathways in later‐life behavioral symptoms. In addition, we performed exploratory MBI domain‐level analyses.

2. METHODS

2.1. Study participants

Data used in the preparation of this article were obtained from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database. ²⁶ ADNI was launched in 2003 as a public‐private partnership, led by Principal Investigator Michael W. Weiner, MD. The primary goal of ADNI has been to test whether serial magnetic resonance imaging, positron emission tomography (PET), other biological markers, and clinical and neuropsychological assessment can be combined to measure the progression of MCI and early Alzheimer's disease.

We included 1273 participants (708 NC, 565 MCI). Cognitive status was determined based on ADNI inclusion criteria. NC participants had normal memory function on the education‐adjusted Logical Memory II subscale of the Wechsler Memory Scale—Revised, ²⁷ , ²⁸ Mini‐Mental State Examination (MMSE) ²⁹ score of 24 to 30, Clinical Dementia Rating ³⁰ and Memory Box scores ³¹ of 0, no significant impairment in daily cognitive functioning, and no subjective memory complaints reported by participant, study partner, or clinician. MCI participants met the criteria for abnormal memory function on the education‐adjusted Logical Memory II subscale, MMSE score of 24 to 30, Clinical Dementia Rating and Memory Box scores ≥ 0.5, preserved general cognition insufficient for Alzheimer's disease diagnosis, and present subjective memory complaints.

The demographic and clinical data collected included age, sex, years of education, Montreal Cognitive Assessment (MoCA), ³² and Neuropsychiatric Inventory (NPI), ³³ an instrument for evaluating behavioral and psychological symptoms by study partner. Additionally, data on handedness, history of hypertension, smoking status, body mass index, plasma amyloid beta (Aβ)42, and plasma phosphorylated tau at threonine 217 (p‐tau217) were collected when available. Participants underwent neuroimaging examinations as part of the ADNI protocol, including high‐resolution T1‐weighted MP‐RAGE sequences and T2‐weighted fluid‐attenuated inversion recovery (FLAIR) sequences on 3T magnetic resonance imaging (MRI) scanners. When multiple eligible pairs of NPI assessment and MRI acquisition were available for a participant, visits were paired to minimize the time difference between behavioral assessment and imaging. The mean ± standard deviation of absolute time difference between the NPI assessment and MRI acquisition used in the analysis was 35 ± 42 days.

RESEARCH IN CONTEXT

1. Systematic review: The authors reviewed the relevant literature using PubMed and Google Scholar. Identified studies investigating WMHs in relation to MBI focused on total WMH volume and included participants with either NC or MCI. No study investigating spatial distribution of WMHs in MBI individuals was found.

2. Interpretation: In a cohort of 1273 participants, we found that greater WMH volume was associated with higher MBI total score and status in NC and MCI participants. Furthermore, higher MBI total scores were associated with increased WMH volume in the left inferior longitudinal fasciculus. These results suggest that WMHs may contribute to later‐life behavioral symptoms.

3. Future directions: Future studies should aim to analyze longitudinal changes of WMH volume and their spatial distribution and whether the progression of WMHs is associated with the onset of MBI symptoms.

Participant selection flowchart is shown in Figure 1. From a total of 2779 participants with the required demographic and clinical data, we excluded participants with major neurocognitive disorders (n = 566, where n is the number of individuals), insufficient data to determine MBI status (n = 490), and corrupted images (n = 51), leaving 1273 individuals in the final analysis sample.

FIGURE 1.

Detailed participant selection flowchart. ADNI, Alzheimer's Disease Neuroimaging Initiative; FLAIR, fluid‐attenuated inversion recovery; MBI, mild behavioral impairment; MCI, mild cognitive impairment; MoCA, Montreal Cognitive Assessment; MP‐RAGE, magnetization prepared rapid gradient‐echo; NC, normal cognition; NPI, Neuropsychiatric Inventory.

2.2. MBI status assessment

Both NC and MCI participants were divided into two groups according to the presence or absence of MBI syndrome. As the MBI Checklist ² is not available in the ADNI database, MBI status was assessed using NPI score transformation. ¹ , ³⁴ First, the subdivision of 10 NPI domains into five MBI domains was performed as follows:

• MBI absence of motivation/drive (NPI apathy/indifference domain)

• MBI emotional dysregulation (NPI depression, anxiety, and elation/euphoria domains)

• MBI impulse dyscontrol (NPI irritability, agitation/aggression, and aberrant motor behavior domains)

• MBI social inappropriateness (NPI disinhibition domain)

• MBI abnormal thoughts/perception (NPI delusions and hallucinations domains)

The scores of each MBI domain were obtained as the sum of products of frequency and severity scores of the corresponding NPI domains, yielding a theoretical range of 0 to 120. The NPI assessment has a 1‐month reference range, and the MBI criteria require a substantial change in ordinary behavior or personality to remain persistent for at least 6 months. Participants were classified as MBI+ if their total MBI score was ≥2 at two consecutive visits at least 6 months apart and as MBI− otherwise. ³⁵

2.3. Processing of T1‐weighted images

The preprocessing of structural T1‐weighted MP‐RAGE images was carried out in the statistical parameter mapping software (SPM12) and MATLAB 2024b. Raw DICOM images were converted to NIfTI format before registration to the standard Montreal Neurological Institute (MNI) space. Anatomical images were segmented to tissue probabilistic maps, corrected for field distortions and non‐uniformity. Total intracranial volume (TIV) was calculated during this stage. Subsequently, a geodesic shooting algorithm was used to diffeomorphically warp tissue probability maps to the MNI standard spatial space template implemented in SPM12. During each preprocessing stage, images were visually inspected to avoid registration errors.

2.4. WMH segmentation

Raw T2‐weighted FLAIR images were registered to original T1‐weighted images before applying the lesion prediction algorithm ³⁶ , ³⁷ as implemented in the lesion segmentation tool version 3.0.0 ³⁸ for SPM12, providing a lesion probability estimate for each voxel. The best correspondence between computational WMH segmentation and the expert's WMH determination was assessed via a probability threshold ranging from 0.05 to 0.95. For each probability threshold value, the Dice similarity coefficient was calculated for 10 participants using reference segmentations of FLAIR images performed by an expert. A threshold of 0.35 (Dice's similarity coefficient = 0.667, Figure S1) was chosen, which is consistent with the literature. ³⁹ , ⁴⁰ Resulting probabilistic maps were brought to the standard MNI152 space using deformation fields generated during preprocessing of T1‐weighted images. Images were subjected to visual inspection after each processing stage.

2.5. Statistical analysis

Demographic and clinical characteristics were presented as mean (sample standard deviation) and number (%) for categorical variables. Normality was assessed using the Jarque‐Bera test. Group comparisons were performed using independent t‐tests or Mann‐Whitney U tests for continuous variables, as appropriate, and chi‐squared tests for categorical variables.

Associations between MBI total score, MBI status, and individual MBI domain‐level scores (independent variables), and WMH volume (dependent variable) was examined using separate general linear models (GLMs), adjusted for age, sex, years of education, and TIV. The addition of MoCA score as a covariate was evaluated, as previously done with MMSE score. ²⁴ Although a causal diagram in the DAGitty software version 3.1 indicated that MoCA adjustment was unnecessary (Figure S2), a likelihood ratio test supported its inclusion; therefore, models with MoCA were also reported. The primary volumetric analyses were defined as the four GLMs examining the associations of MBI total score and MBI status with WMH volume in NC and MCI participants, without adjustment for MoCA score. To account for multiple testing across these primary volumetric analyses, Benjamini‐Hochberg false discovery rate correction was applied across the four primary tests. Models additionally adjusted for MoCA score were considered in additional adjustment analyses of the same associations, and domain‐level analyses were considered exploratory given the lower endorsement of individual MBI domain scores. The abnormal thoughts/perception domain was excluded from the domain‐level analysis due to its low prevalence. Moreover, WMH volumes were natural log‐transformed to reduce right skewness of residuals. To visualize the association between MBI total score and total WMH volume, we generated an adjusted scatterplot with fitted regression lines for NC and MCI participants. The fitted values and 95% confidence intervals were derived from the corresponding GLM adjusted for age, sex, years of education, and TIV.

To address the potential influence of vascular risk factors and Alzheimer's disease‐related plasma biomarkers, we additionally performed an exploratory complete‐case sensitivity analysis in participants with available vascular risk factors ⁴¹ (history of hypertension, smoking status, and body mass index), plasma Aβ42, and plasma p‐tau217. Since complete data were available only in a much smaller subset of participants (n = 241), NC and MCI participants were analyzed together with diagnostic group included as an additional covariate. These models were adjusted for age, sex, years of education, diagnostic group, TIV, vascular risk factors, and plasma biomarkers. Interaction terms were not included since these models were considered insufficiently powered for reliable interaction testing.

Voxel‐wise multiple regression analyses were conducted in SPM12 on the whole cohort (n = 1273) to examine associations between WMH spatial distribution and MBI status or total score. Analyses were restricted to voxels affected by WMHs in at least 50 participants using an explicit mask in MNI space. Thresholded WMH maps were smoothed with a 6‐mm full‐width at half‐maximum Gaussian kernel and analyzed using a voxel‐wise GLM with either MBI status or MBI total score along with age, sex, years of education, TIV, and MCI versus NC status as covariates. ⁴² Clusters surviving family‐wise error (FWE) correction (p  < 0.001, cluster level) with a minimum extent of 150 voxels (0.5 mL) were considered significant. Given the lateralized nature of significant findings, we additionally performed a sensitivity analysis in participants with available dominant hand information (1228 individuals, 10.8% left‐handed), adding handedness as an additional covariate to the voxel‐wise model. Robustness was assessed using non‐parametric permutation testing. All analyses were performed in MATLAB 2024b.

3. RESULTS

3.1. Demographic characteristics

The demographic characteristics of the participants are summarized in Table 1. The participants were between 50 and 95 years of age. In total, 247 MBI+ participants were left after selection across both cognitive groups. In both cognitive groups, MBI+ individuals were older, had lower MoCA scores, and had both greater TIV and WMH volume. Furthermore, a smaller proportion of females were present in both MBI+ subgroups. In NC participants, the observed ranges were 0 to 27 for NPI total score, 0 to 27 for MBI total score, 0 to 8 for motivation/drive, 0 to 13 for emotional dysregulation, 0 to 12 for impulse control, 0 to 3 for social inappropriateness, and none for abnormal thoughts. In MCI participants, the observed ranges were 2 to 78 for NPI total score, 1 to 54 for MBI total score, 0 to 12 for motivation/drive, 0 to 20 for emotional dysregulation, 0 to 21 for impulse control, 0 to 12 for social inappropriateness, and 0 to 10 for abnormal thoughts.

TABLE 1.

Characteristics of participants.

Characteristics	NC		p value	MCI		p value
	MBI− (n = 643)	MBI+ (n = 65)		MBI− (n = 383)	MBI+ (n = 182)
Age (years)	72.0 (8.06)	74.6 (6.35)	0.01	72.8 (7.65)	74.1 (7.86)	0.06
Female sex	385 (59.9%)	27 (41.5%)	<0.01	205 (53.5%)	70 (38.5%)	<0.01
Education (years)	16.7 (2.45)	16.5 (2.14)	0.35	16.2 (2.64)	16.0 (2.71)	0.65
MoCA score	26.1 (2.56)	25.3 (2.64)	0.01	23.3 (3.64)	22.9 (3.56)	0.25
TIV (dm3)	1.47 (0.15)	1.54 (0.14)	<0.01	1.47 (0.15)	1.53 (0.15)	<0.01
NPI score	1.04 (2.88)	6.40 (4.68)	<0.01	2.67 (5.35)	10.3 (9.34)	<0.01
Motivation/drive	0.06 (0.59)	0.58 (1.39)	<0.01	0.27 (1.08)	1.40 (2.32)	<0.01
Emotional dysregulation	0.27 (1.09)	1.69 (2.23)	<0.01	0.62 (1.92)	2.16 (2.70)	<0.01
Impulse dyscontrol	0.20 (0.97)	2.58 (2.55)	<0.01	0.57 (1.78)	3.07 (3.73)	<0.01
Social inappropriateness	0.03 (0.25)	0.20 (0.56)	<0.01	0.13 (0.62)	0.74 (1.83)	<0.01
Abnormal thoughts	0.00 (0.00)	0.00 (0.00)	–	0.03 (0.33)	0.16 (0.94)	<0.01
MBI total score	0.57 (2.23)	5.06 (3.45)	<0.01	1.62 (4.11)	7.55 (7.17)	<0.01
WMH volume (cm3)	9.94 (13.1)	13.5 (14.2)	<0.01	13.9 (15.0)	19.1 (19.6)	<0.01

Note: Categorical variables were compared via chi‐squared test and reported with n (%); continuous variables are reported as mean (sample standard deviation); normally distributed continuous variables were compared via t‐test and non‐normally distributed continuous variables were compared via u‐test.

Abbreviations: MBI, mild behavioral impairment; MoCA, Montreal Cognitive Assessment; NPI, Neuropsychiatric Inventory; TIV, total intracranial volume; WMH, white matter hyperintensity.

3.2. Associations in NC participants

A higher total MBI score was significantly associated with greater WMH volume without (estimate: 2.62%, p = 0.019) and with (estimate: 2.44%, p  = 0.028) MoCA adjustment (Table 2). MBI status was significantly associated with greater WMH volume without MoCA adjustment (estimate: 22.5%, p  = 0.049) but not in the GLM with MoCA adjustment (estimate: 20.8%, p  = 0.067). No individual MBI domain‐level score was significantly associated with WMH volume in NC participants. In both cases with and without adjustment for the MoCA score, age, female sex, and TIV were consistently associated with higher WMH volume. Additionally, in the GLM adjusted for MoCA score, a lower MoCA score was associated with higher WMH volume (estimate: −2.53%, p  = 0.031 in Table S1). Each GLM with covariates can be found in Tables S1–S6.

TABLE 2.

Associations between MBI measures and WMH volume in participants with normal cognition with and without adjustment for MoCA score.

Model	Parameters	Estimate (%)	95% CI (%)			p value
Without MoCA adjustment	MBI total score	2.62	0.43	—	4.85	0.019
	MBI+ versus MBI−	22.5	0.02	—	49.9	0.049
	Motivation/drive	−3.08	−11.2	—	5.90	0.488
	Emotional dysregulation	4.72	−0.58	—	10.3	0.082
	Impulse dyscontrol	3.64	−1.10	—	8.62	0.134
	Social inappropriateness	−4.94	−22.5	—	16.5	0.625
With MoCA Adjustment	MBI total score	2.44	0.26	—	4.67	0.028
	MBI+ versus MBI−	20.8	−1.31	—	47.8	0.067
	Motivation/drive	−3.42	−11.6	—	5.50	0.440
	Emotional dysregulation	4.77	−0.52	—	10.4	0.078
	Impulse dyscontrol	3.26	−1.47	—	8.21	0.179
	Social inappropriateness	−4.73	−22.2	—	16.8	0.639

Note: White matter hyperintensity volumes were natural log‐transformed. MBI measures (total score, status, domain‐level scores) were tested in independent models. Each model was adjusted for age, sex, years of education, and TIV. TIV was entered in milliliters.

Abbreviations: CI, confidence interval; MBI, mild behavioral impairment; MoCA, Montreal Cognitive Assessment; WMH, white matter hyperintensity.

3.3. Associations in MCI participants

In the MCI participants, MBI total score was significantly associated with greater WMH volume without (estimate: 1.38%, p  = 0.016) and with (estimate: 1.29%, p  = 0.024) adjustment for MoCA score (Table 3). Likewise, MBI status was significantly associated in both GLM without (estimate: 19.8%, p  = 0.014) and with (estimate: 19.3%, p  = 0.015) MoCA adjustment. Across the four primary GLMs, including MBI total score and MBI status in NC and MCI participants without MoCA adjustment, all associations remained significant after Benjamini‐Hochberg false discovery rate correction (range 0.025 to 0.049). No individual MBI domain‐level score was significantly associated with WMH volume in MCI participants. Among the covariates, increasing age, female sex, and larger TIV were significantly associated with a higher WMH volume. In GLM adjusted for MoCA, a lower MoCA score was associated with higher WMH volume (estimate: −2.24%, p  = 0.018 in Table S7). Each GLM with covariates can be found in Tables S7–S12. Figure 2a visualizes these group‐specific associations in NC and MCI participants.

TABLE 3.

Associations between MBI measures and WMH volume in participants with mild cognitive impairment with and without adjustment for MoCA score.

Model	Parameters	Estimate (%)	95% CI (%)			p value
Without MoCA Adjustment	MBI total score	1.38	0.26	—	2.52	0.016
	MBI+ versus MBI−	19.8	3.78	—	38.2	0.014
	Motivation/drive	2.51	−1.94	—	7.17	0.273
	Emotional dysregulation	2.45	−0.72	—	5.72	0.131
	Impulse dyscontrol	0.06	−2.68	—	2.88	0.964
	Social inappropriateness	0.85	−5.54	—	7.67	0.799
With MoCA Adjustment	MBI total score	1.29	0.17	—	2.43	0.024
	MBI+ versus MBI−	19.3	3.47	—	37.6	0.015
	Motivation/drive	2.32	−2.11	—	6.94	0.309
	Emotional dysregulation	2.51	−0.64	—	5.77	0.119
	Impulse dyscontrol	−0.23	−2.96	—	2.58	0.872
	Social inappropriateness	1.19	−5.19	—	8.00	0.721

Note: White matter hyperintensity volumes were natural log‐transformed. MBI measures (total score, status, domain‐level scores) were tested in independent models. Each model was adjusted for age, sex, years of education, and TIV. TIV was entered in milliliters here.

Abbreviations: CI, confidence interval; MBI, mild behavioral impairment; MoCA, Montreal Cognitive Assessment; WMH, white matter hyperintensity.

FIGURE 2.

Association between MBI total score and total WMH volume, and significant cluster positively associated with MBI total score in voxel‐wise analysis. (A) Association between MBI total score and total WMH volume for participants with NC and MCI. Blue circles: NC participants, red circles: MCI participants, fitted lines and shaded areas: adjusted associations and 95% confidence intervals derived from models adjusted for age, sex, years of education, and TIV. (B–D) Various perspectives of identified cluster: (B) posterior, (C) left, (D) inferior. Model in voxel‐wise analysis was adjusted for age, sex, education, TIV, and MCI versus NC status. Cluster survived voxel‐wise FWE‐correction pFWE < 0.001, k ≥ 150 voxels. FWE, family‐wise error; MBI, mild behavioral impairment; MCI, mild cognitive impairment; NC, normal cognition; TIV, total intracranial volume; WMH, white matter hyperintensity.

In an exploratory complete‐case sensitivity analysis, we evaluated whether adjustment for vascular risk factors and plasma biomarkers altered the association between MBI measures and WMH volume. Complete data on hypertension, smoking status, body mass index, plasma Aβ42, and plasma p‐tau217 were available for 241 participants. In this restricted subset, MBI total score and MBI status were not significantly associated with WMH volume in the basic models adjusted for age, sex, education, diagnostic group, and TIV. The associations remained non‐significant after additional adjustment for vascular risk factors and plasma biomarkers (Tables S13 and S14).

3.4. Voxel‐wise regression analyses

No significant clusters were identified when comparing the MBI+ and MBI− participants. However, higher total MBI scores were significantly associated with one significant cluster, comprising three distinct peaks, located primarily (80%) in the left inferior longitudinal fasciculus and, to a lesser extent (both < 10%), in the left inferior fronto‐occipital fasciculus and the posterior segment of the left arcuate fasciculus (Table 4, Figure 2b–d). Results from SPM12 can be found in Figure S3. The handedness‐adjusted sensitivity analysis showed the same left‐hemispheric cluster with an identical main peak (−38, −40, −10; Figure S4). Cluster‐level robustness was further supported by a permutation test of extracted cluster WMH volume and MBI total score (Spearman ρ = 0.142, p < 0.001; 10,000 permutations), where none of the permutations exceeded the original correlation in absolute terms.

TABLE 4.

Significant cluster positively associated with MBI total score in voxel‐wise analysis of WMH probability maps.

		MNI coordinates			Volume (mm3)	T value	Cluster P FWE	Peak P FWE
Peak label	Peak brain structure	x	y	z
Peak 1 (main)	Left inferior longitudinal fasciculus	−38	−40	−10	1238	7.14	<0.001	<0.001
Peak 2	Left inferior fronto‐occipital fasciculus	−34	−51	−6		5.54	<0.001	<0.001
Peak 3	Left inferior longitudinal fasciculus	−42	−39	4		5.54	<0.001	<0.001

Note: Model was adjusted for age, sex, education, TIV, and MCI versus NC status. Cluster survived voxel‐wise FWE‐correction P _FWE < 0.001, k ≥ 150 voxels.

Abbreviations: FWE, family‐wise error; MBI, mild behavioral impairment; MCI, mild cognitive impairment; MNI, Montreal Neurological Institute; NC, normal cognition; TIV, total intracranial volume; WMH, white matter hyperintensity.

4. DISCUSSION

In this study, we found that higher MBI total score and MBI status were associated with greater WMH volume in both NC and MCI participants. Our exploratory analysis did not show any specific domain of MBI that would explain this association. Voxel‐wise analysis revealed that higher MBI total score was associated with higher probability of WMH lesions in the left inferior longitudinal fasciculus, extending to the left inferior fronto‐occipital and arcuate fasciculi. These findings suggest a possible contribution of WMH volume and spatial distribution to later‐life behavioral symptoms.

Our findings are consistent with studies that have examined the relationship between MBI and WMHs. Miao et al. ²⁴ found that MBI status, rather than individual MBI domain‐level scores, was associated with WMH volume in a cohort of MCI participants, suggesting that MBI may serve as an early marker of individuals with greater WMH volume and at risk of developing dementia. Similarly, a study by Yang et al. ²⁵ linked gray matter atrophy and reduced cortical thickness related to WMHs to MBI total score in cognitively normal individuals. WMHs have also been associated with incident depression and worsening NPI scores. ¹² , ⁴³ Furthermore, Li et al. ¹¹ associated cerebral small vessel disease with more severe anxiety, depression, and other neuropsychiatric symptoms, exacerbated by poor sleep quality. Our results extend this literature by showing that WMH volume and spatial distribution are associated with MBI status and total score in both NC and MCI participants, while individual domains were not significant, emphasizing the importance of MBI as a global syndrome over individual domains or symptoms.

The inferior longitudinal fasciculus is a multifunctional associative white matter tract connecting occipito‐temporal regions with the anterior temporal regions. It has been implicated in object and face recognition, visual memory, reading, lexical and semantic processing, proper name retrieval, and emotional processing. Although the inferior longitudinal fasciculus is bilateral, some functions show hemispheric specialization. Left‐sided involvement has been linked particularly to reading, lexical and semantic processing, proper name retrieval, and emotion recognition. ⁴⁴ , ⁴⁵ In the context of our findings, WMH lesions in the left inferior longitudinal fasciculus may therefore provide a plausible neuroanatomical link between regional WMH‐associated white matter changes and behavioral or affective symptoms captured by MBI, although compensation by other structures or the contralateral tract may occur. Importantly, the handedness‐adjusted sensitivity analysis showed a stable spatial pattern and identical peak location, suggesting that the observed lateralized association was not driven by variation in hand dominance.

Here, we applied a cut‐off of 2 points to classify MBI+ because the NPI score transformation involved multiplying domain frequency and severity, and MBI+ status required two consecutive visits meeting this criterion. ³⁵ This resulted in an overall MBI prevalence of 19.4%, lower than previous NPI/NPI‐Questionnaire‐based estimates of 26.4% to 32.3% in NC and/or MCI cohorts. ³⁵ , ⁴⁶ This difference likely reflects participant selection and the stricter cut‐off used in our study compared with the traditional 1‐point threshold used with NPI‐Questionnaire scores.

In GLM analyses, NC and MCI participants were examined separately to assess whether associations between MBI measures and WMH volume differed by diagnostic group. MoCA‐adjusted models were considered additional adjustment analyses of the same associations rather than primary tests. Adding MoCA slightly reduced the estimates for MBI measures. In NC participants, MBI status became non‐significant after MoCA adjustment, although the estimate remained similar (decrease from 22.5% to 20.8%), suggesting that the association was not largely explained by global cognitive performance. In MCI participants, MoCA adjustment had little effect on the estimates or significance levels of MBI measures. Together, these findings suggest that the observed associations between MBI and WMH volume are not solely determined by cognitive status. Voxel‐wise analyses were performed in the whole cohort because MBI can occur in both NC and MCI individuals, with diagnostic group included as a covariate. Although analyses restricted to NC participants did not reveal clusters of sufficient size, the whole‐cohort analysis supported and improved the robustness of the observed results.

Although this study focused on WMH, MBI has also been associated with amyloid, p‐tau, neurofilament light, and regional brain atrophy. ⁴ , ⁶ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ , ²¹ , ²² , ²³ Thus, our findings suggest that WMH‐associated white matter damage may contribute to MBI symptoms, but they do not establish a specific vascular mechanism. Exploratory complete‐case analyses including vascular risk factors and plasma biomarkers were limited to 241 participants. In this subset, MBI measures were not significantly associated with WMHs even before additional adjustment. These null findings likely reflect reduced power and complete‐case selection rather than attenuation by vascular or Alzheimer's disease‐related biomarkers. Future larger multimodal studies are needed to test interactions among WMHs, vascular risk factors, amyloid, p‐tau, and MBI.

A key strength of this study is the use of a large and well‐characterized cohort of 1273 participants from the ADNI database, which provided sufficient statistical power to investigate the relationship between MBI and WMHs. Nevertheless, several limitations should be considered. MBI was derived from NPI rather than from MBI Checklist, as the MBI Checklist is not available in the ADNI. In addition, domain‐level analyses were exploratory and may have been underpowered because individual symptom domains showed limited endorsement and restricted score ranges. Therefore, the lack of significant domain‐level findings should not be interpreted as evidence that specific MBI domains are unrelated to WMHs. Automatic segmentation algorithms may overestimate lesion volume, although we employed a validated lesion prediction algorithm and an empirically selected threshold. ³⁹ , ⁴⁰

In conclusion, MBI total score and status were associated with greater WMH volume in dementia‐free individuals, and higher MBI total score was associated with WMHs in the left inferior longitudinal fasciculus. These findings support an association between WMH‐associated white matter damage and later‐life behavioral symptoms, while emphasizing the need for longitudinal and multimodal studies using MBI Checklist data.

Further studies should also focus on longitudinal monitoring of WMHs in relation to the development of MBI and on changes in specific MBI domains using MBI Checklist data, which may provide greater sensitivity for domain‐level analyses than the NPI‐derived domain scores.

AUTHOR CONTRIBUTIONS

ZL contributed to the study conception and design, data curation and processing, development and implementation of the statistical analyses, data visualization, interpretation of the results, and drafting of the manuscript. RD contributed to the methodological design, development of the data‐processing workflow, and design of the statistical analyses. MS contributed to the study design, interpretation of the results, and manuscript revision. PBK contributed to the study design, interpretation of the results, manuscript drafting, and critical revision. ZI contributed methodological and clinical expertise, interpretation of the results, and critical revision of the manuscript. The Alzheimer's Disease Neuroimaging Initiative (ADNI) investigators contributed to the design and implementation of ADNI and/or provided data but did not participate in the analysis or writing of this report. All individual authors reviewed and approved the final version of the manuscript.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest. Author disclosures are available in the Supporting Information.

CONSENT STATEMENT

All participants in the ADNI provided written informed consent at enrollment. This study is a retrospective analysis of de‐identified ADNI data; therefore, no additional informed consent was required.

Supporting information

Supporting Information

Supporting Information

Linha Z, Dolezal R, Seifert M, Ravi T, Ismail Z, Kearns PB. Associations of mild behavioral impairment and white matter hyperintensity patterns. Alzheimer's Dement. 2026;18:e70500. 10.1002/dad2.70500