Are neuropsychiatric symptoms a marker of small vessel disease progression in older adults? Evidence from the Lothian Birth Cohort 1936

Abstract Background Neuropsychiatric symptoms could form part of an early cerebral small vessel disease prodrome that is detectable before stroke or dementia onset. We aimed to identify whether apathy, depression, anxiety, and subjective memory complaints associate with longitudinal white matter hyperintensity (WMH) progression. Methods Community‐dwelling older adults from the observational Lothian Birth Cohort 1936 attended three visits at mean ages 73, 76, and 79 years, repeating MRI, Mini‐Mental State Examination, neuropsychiatric (Dimensional Apathy Scale, Hospital Anxiety and Depression Scale), and subjective memory symptoms. We ran regression and mixed‐effects models for symptoms and normalised WMH volumes (cube root of WMH:ICV × 10). Results At age 73, 76, and 79, m = 672, n = 476, and n = 382 participants attended MRI respectively. Worse apathy at age 79 was associated with WMH volume increase (β = 0.27, p = 0.04) in the preceding 6 years. A 1SD increase in apathy score at age 79 associated with a 0.17 increase in WMH (β = 0.17 normalised WMH percent ICV, p = 0.009). In apathy subscales, executive (β = 0.13, p = 0.05) and emotional (β = 0.13, p = 0.04) scores associated with increasing WMH more than initiation scores (β = 0.11, p = 0.08). Increasing WMH also associated with age (β = 0.40, p = 0.002) but not higher depression (β = ‐0.01, p = 0.78), anxiety (β = 0.05, p = 0.13) scores, or subjective memory complaints (β = 1.12, p = 0.75). Conclusions Apathy independently associates with preceding longitudinal WMH progression, while depression, anxiety, and subjective memory complaints do not. Patients with apathy should be considered for enrolment to small vessel disease trials.


| INTRODUCTION
Cerebral small vessel disease (SVD) is highly prevalent in older people and is a common cause of vascular dementia and stroke. 1 It is not clear whether early symptoms of SVD could predict the development of these conditions. Identifying an 'SVD syndrome' would allow clinicians and researchers to identify individuals who might benefit from future clinical trials or earlier SVD preventative interventions. 2 Neuropsychiatric and cognitive symptoms have been proposed as potential candidates for early clinical detection of SVD. Recent work has established that cross-sectional relationships exist between these symptoms and worse white matter hyperintensity (WMH) burden, 3,4 a key radiological feature of SVD. 5 Specifically, more severe WMH burden associates with apathy and depression. In contrast, more severe WMH burden does not associate cross-sectionally with subjective memory complaints (SMCs), while associations with anxiety and other neuropsychiatric symptoms are indeterminate. 3,4 Longitudinal imaging-symptoms studies are required to confirm these associations.
However, such studies are sparse (online supplemental Table 1). [6][7][8][9][10][11] We aimed to determine whether anxiety, depression, apathy, or SMCs are each associated with longitudinal WMH volume change in an older community-dwelling cohort over 6 years. respectively. Participants were invited to MRI scans, which were first conducted at Wave two. The present analysis focuses on Waves two to four, that is, individuals at mean ages of 73, 76, and 79 years at study visits. The protocol 12 and population profiles are published in detail. 13,14 Figure 1 shows the study flow chart.

| Participants
The cognitive status of the LBC1936 has been described previously and applies to 397/476 (83%) of Wave three and 319/382 (84%) of Wave four participants included in the present analysis. 15,16 The majority did not have cognitive impairment and none of the participants reported a dementia diagnosis at recruitment. Three participants had had a dementia diagnosis by Wave three. In participants with relevant data available to make a mild cognitive impairment (MCI) diagnosis, 17 MCI was present in 15% at Wave three and 17% at Wave four. 15

| Neuropsychiatric and cognitive symptom assessments
All symptoms were assessed using self-completed questionnaires.
Depression and anxiety were assessed using the Hospital Anxiety and Depression Scale 18 at each wave (score ranges 0-21).
Apathy was assessed using the Dimensional Apathy Scale (DAS) 19 at age 79 (Wave four) only. The DAS is a multidimensional scale which assesses different apathy subtypes 19 based on subtype relationships to damage and impaired connectivity in the pre-frontal cortico-subcortical and basal ganglia regions. 20 Subscale analysis of the DAS provides further insight into global apathy by looking at specific characteristics of demotivation as follows. Executive apathy is a lack of motivation for planning, organisation and attention.
Emotional apathy is a lack of emotional motivation, indifference, emotional neutrality, blunting or flatness. Initiation apathy is a lack of motivation for self-generated thoughts and/or behaviours. The total score ranges from 0 to 72 and the subscales from 0 to 24. Although there is no predefined cut-off for the DAS, normative data have previously been used to suggest abnormality level cut-offs for each subscale based on ≥2 SD above the mean. 21 We did not use a cut-off for the present analysis. Higher scores indicate worse apathy.
The present analysis includes three questions about SMCs which were asked at age 79 (Wave four) only: Q1 "Do you currently have any problems with your memory?", Q2 "If yes, are these problems interfering with your normal life?", and Q3 "Do you forget where you have left things more often than you used to?". We describe each symptom individually but use Question 1 in multivariable analyses, to maintain consistency with the existing SMC literature.

| Clinical assessments
Self-reported years of education, that is, full-time schooling, were collected at Wave one. The following measures were collected at all waves. Self-reported vascular risk factor diagnoses were recorded, that is, binary history (yes/no) of cardiovascular disease, diabetes, hypercholesterolaemia, hypertension, stroke history, and smoking status (current/ex/never). We summed a composite vascular risk factor score containing these variables, described previously, 22

| MRI analysis
We combined well-validated computational methods with manual checking to quantify WMH volumes, per the Standards for ReportIng Vascular Changes on Neuroimaging (STRIVE) guidelines. 5 Quantitative analysis methods are detailed elsewhere. 28,29 In brief, we used a semi-automated multispectral segmentation tool to extract volumes including WMH and intracranial volume (ICV), co-registering T1-, T2-, T2*-weighted and FLAIR images to the T2-weighted images using FSL-FLIRT. This highly reproducible method, complemented by manually checking all images, includes distinction between WMH and chronic infarcts and is overseen by an experienced neuroradiologist (JMW). Image analysts and neuroradiologists were blinded to all nonimaging data. Full structural brain characteristics for the cohort are

| Statistical analysis
We compared participants who did or did not attend (a) the first MRI visit and (b) follow-up MRI visits. We expressed WMH volume as a ratio of ICV and calculated the cube root to optimise model fit and F I G U R E 1 Flow chart of study participants. CLANCY ET AL.
-3 of 11 minimise model complexity, as described previously, 31 multiplying all values by 10 to scale with other variables. We created a sum vascular risk factor score comprising hypertension, hyperlipidaemia, smoking status, and diabetes, assigning equal weight to the presence of each factor, as described previously. 32 The rationale for this was to include relevant variables while avoiding model overfitting. For descriptive symptom analyses, we calculated Pearson (r) and Spearman (rs) correlation coefficients and Wilcoxon rank-sum tests (W).
We assessed associations between WMH and each neuropsychiatric or cognitive symptom as a dependent variable to explore whether SVD lesions associate with neuropsychiatric and cognitive symptoms. We describe data collection timelines in Figure 1.
First, to assess cross-sectional symptom-WMH associations at each wave, we ran separate linear regression models using Wave two (anxiety, depression), three (anxiety, depression), and four (anxiety, depression, apathy, SMCs) data. Then, to assess longitudinal anxiety and depression associations with WMH, we ran separate linear mixed-effects models with each symptom as the outcome and normalised WMH volume (expressed as cube root of WMH:ICV [�10]) as the exposure, using data from all three time-points. We adjusted for time (i.e. wave number), age at each scan, MMSE, sex, anxiety/ depression, vascular risk factors, and years of education. We additionally adjusted for disability in depression models to account for depression-disability associations. 33 Since apathy and SMCs were collected at age 79 (Wave four) only, the longitudinal assessment analysis of these symptoms required a different statistical approach which assessed preceding WMH volume change. For each model, we defined WMH volume change as the difference between age 73 (Wave two) and age 79 (Wave four) WMH volumes. We ran linear regression (apathy) and generalised linear models (SMCs) with each symptom as the outcome and assessed associations with WMH volume change. We corrected for baseline WMH volumes, MMSE, depression scores, activities of daily living, vascular risk factor scores, age, sex, and education.
In our final set of analyses we reversed the research question, using WMH as the outcome and neuropsychiatric symptoms as the predictors. All neuropsychiatric symptoms (apathy, subjective memory complaints, anxiety, depression) were included as predictors in this final model (online supplemental Figure 8).
Since longitudinal research assessing neuropsychiatric symptoms and SVD progression is a relatively new area of exploration in the SVD-neuropsychiatric symptom literature, 3 we regard this as an exploratory analysis. Therefore, to avoid the probability of a Type II error, 34 we planned all comparisons prior to data analysis, set alpha level to 0.05, reported all analyses performed, and did not correct for multiple comparisons so that any given p value can be interpreted based on the analysis in question. 35 We present fully adjusted models defined according to our clinical hypothesis, adjusted for all clinically relevant covariates supported by previous literature.
We report standardised beta (β) coefficients. For all mixedeffects models, we fitted individual participants as the random effect to assess variation within individuals, using R package lme4. 36

| Baseline population characteristics
We include n = 672 participants who attended MRI at mean age 73 (Wave two), n = 476 at mean age 76 (Wave three), and n = 382 at mean age 79 (Wave four) (Figure 1). At Wave two, there were no differences in age, sex, vascular risk factor scores, disability, MMSE, anxiety, or depression among participants who did (n = 672) versus did not (n = 193) take part in the optional MRI sub-study (online supplemental    Table 1 shows population characteristics at each wave.

| WMH change as predictor and wave 2/3/4 anxiety as outcome
In cross-sectional analyses, anxiety scores were not associated with WMH volumes at any wave (at Wave four, β = −0.01, p = 0.82; online supplemental Table 6/online supplemental Figure 4A). Longitudinal mixed-effects analysis (n = 689) produced similar results, that is, anxiety scores were not associated with changing WMH volumes

| WMH change as predictor and wave 4 apathy as outcome
In cross-sectional analysis (n = 197), there was a small trend towards an association between higher apathy scores and WMH volumes at T A B L E 1 Population characteristics of participants who attended MRI at waves two, three, or four    Table 9/ online supplemental Figure 9). Higher scores on the combined emotional plus executive DAS subscales (β = 0.18, p = 0.006) had a stronger association with WMH volume change than initiation plus emotional (β = 0.14, p = 0.02) or initiation plus executive (β = 0.14, p = 0.02) combined subscales. There was no association between vascular risk factor scores and WMH, adjusted for age (Table 3, online supplemental Figure 8). Associations between apathy and WMH increase remained after adjusting for stroke history (online supplemental Figure 8).

| DISCUSSION
The present study indicates that higher apathy scores at age 79 are The LBC1936 is a well-characterised cohort with rich longitudinal data, allowing adjustment for key confounders. The analysis had limitations.We did not assess the full spectrum of SVD-related pathology. Further work is needed to establish how other structural features (e.g. perivascular spaces, lacunes) and subvisible abnormal- ities (e.g. on diffusion imaging) manifest clinically across different locations, stages of development, and severity. 37,38 We focused on total WMH volumes rather than specific regional networks, 39  A one standard deviation increment in apathy score at age 79 was associated with a 0.17 increase in normalised WMH volumes in the preceding 6 years, and at age 79, individuals with the greatest preceding WMH volume increase scored three apathy points higher than individuals who had had the smallest increase in WMH volumes.
This represents a small but potentially important and recognisable association, motivating further work into the role of apathy as a clinical marker of SVD. Depression per se, anxiety, and SMCs require further assessment in a range of populations.

| Future directions for SVD symptom-lesion associations
We have highlighted a key timeframe in older age during which SVD progresses rapidly, with WMH almost doubling over 6 years, representing a vital window of opportunity to target potential treatments.
It is too early to say whether apathy scores would be useful on an individual basis, but further research into clinical prediction models is needed to guide whether treatments could be trialled in individuals who are identified as being at the highest risk for WMH progression, that is, older adults with apathy. Longitudinal apathy analyses would Symptom-lesion associations need to be explored at varying time intervals, assessing acute, subacute, and chronic brain changes. Our research question should also be extended to other non-focal symptoms that are not currently sufficient to contribute to a stroke or dementia diagnosis.
The findings from this study will assist in identifying an early SVD syndrome and encourage us to broaden our focus beyond stroke and dementia. Apathy may help to distinguish SVD 'progressors' from 'non-progressors' in the general population. We need to accelerate efforts to uncover other subtle symptoms of SVD progression. This will identify an as-yet underrepresented population who could benefit from early targeted potential SVD treatments.

AUTHOR CONTRIBUTIONS
Drafting