Vascular disease and apathy symptoms in the very old: A cross‐sectional and longitudinal meta‐analysis of individual participant data

Abstract Objectives Previous findings suggest a vascular foundation underlying apathy, but transdiagnostic and prospective evidence on vascular apathy is scarce. This study examines the association between vascular disease and the presence and development of apathy symptoms in the very old. Methods Four cohorts of the Towards Understanding Longitudinal International older People Studies (TULIPS)‐consortium were included in a two‐staged, individual participant data meta‐analysis using generalized linear mixed models. Vascular disease was defined as a history of any clinical atherosclerotic pathology (angina pectoris, myocardial infarction, intermittent claudication, transient ischemic attack, stroke or related surgeries) and was related to apathy symptoms as repeatedly measured by the Geriatric Depression Scale (GDS‐3A ≥2) over a maximum of 5 years. Results Of all 1868 participants (median age 85 years old), 53.9% had vascular disease and 44.3% experienced apathy symptoms. Participants with vascular disease had a 76% higher risk of apathy symptoms at baseline (odds ratio (OR) 1.76, 95% confidence interval (CI) 1.32–2.35), irrespective of depressive symptoms and only partially explained by stroke. Conversely, there was no association of vascular disease with the occurrence of apathy symptoms longitudinally, both in those with apathy at baseline (OR 1.00, 95% CI 0.84–1.20) and without (OR 0.96, 95% CI 0.84–1.09). Conclusions Vascular disease in the very old is associated with apathy symptoms cross‐sectionally, but not proven longitudinally, independent of depressive symptoms. These findings query a vascular cause underlying apathy symptoms. However, the consistency of our cross‐sectional findings in direction and magnitude across the TULIPS‐consortium do emphasize international relevance of the interplay of vascular factors and apathy in advanced age, which meaning needs further unravelling.


| INTRODUCTION
Reduced motivation, or apathy, is defined as a persistent lack of goaldirected behaviour which impairs daily functioning. 1,2 Apathy symptoms do not merely occur with different medical conditions like dementia or depression, [3][4][5][6][7] but are also common in the general population at old age. These symptoms affect community-dwelling older adults in approximately 20%-40% and occur in the absence of depressive symptoms (i.e. isolated apathy) in 13.8% on average. 8 Due to this widespread occurrence and little evidence about apathy prevention or recovery, 7,9,10 there is a strong need for a transdiagnostic approach to unravel the complex pathophysiology underlying these symptoms. 9 Vascular pathologies have been proposed to specifically precede apathy symptoms and not depressive symptoms, and to predispose older adults to developing apathy in old age: the so-called vascular apathy hypothesis. 11 An increasing body of literature has since emerged on the interplay between apathy and several proxies of cardiovascular disease (CVD), as comprehensively depicted in a recent systematic review. 12 Various mechanisms could drive the vascular foundation of apathy, such as atherosclerosis or structural brain damage following stroke. 4,9,11,13 Conversely, apathy symptoms also seem to contribute to incident CVD, 14 which could for example, be mediated by adverse health behaviour if motivation for health promotion declines. 15 To our knowledge, only the latter perspective has been examined over time in a thorough meta-analysis, 8 whereas prospective evidence on the vascular apathy hypothesis remains scarce 12 as well as replication of previous cross-sectional findings in the very old population (i.e. those over 80 years of age), 11 hampering further understanding about the direction and interpretability of these associations in old age. Therefore, we aim to investigate the association between vascular disease and the presence as well as development of apathy symptoms in the very old, irrespective of depressive symptoms.

| Study design and participants
Our two-stage meta-analysis incorporated individual participant data (IPD) from the pre-existing Towards Understanding Longitudinal International older People Studies (TULIPS)-consortium. [16][17][18][19] This is a worldwide collection of prospective, population-based studies on ageing in the very old, which was at the time established to increase power and to examine international variation in associations in old age. We included four of its cohorts from three different studies: (1) 23 All the above final samples were shown to be by and large representative of the several ageing source populations, as outlined in detail, [20][21][22][23] and within these samples individual participants were solely excluded from this study due to data unavailability on the exposure and/or outcome.
Repeated data collection was annual over a maximum of 5 years in the Leiden 85-plus Study as well as LiLACS NZ, while follow-up was clustered at 1.5, 3 and 5 years from baseline in the Newcastle 85+ Study. The individual study protocols were approved by Medical Ethical Committees at their respective sites and all very old or their proxies had granted either oral or written informed consent.

| Vascular disease
Several manifestations of vascular disease have been assessed at baseline by combining different types of data collection per study (Supplementary Table 1). In the Leiden 85-plus Study certified and trained research staff conducted standardized interviews with physicians. 11 In LiLACS NZ such interviews were performed with participants and were seconded by primary health care and hospitalization record checks. 24 In the Newcastle 85+ Study primary health care records were thoroughly examined. 23 Electrocardiograms were recorded in all studies to objectify and to complement findings on previous myocardial infarction. 11,23,24 Vascular disease (yes/no) was defined as a history of any clinical atherosclerotic pathology: angina pectoris, myocardial infarction, intermittent claudication, transient ischemic attack, stroke or any related coronary or peripheral artery surgery.

| Apathy and depressive symptoms
All studies repeatedly administered the 15-item Geriatric Depression Scale (GDS) to their participants, 25 which is appropriate to assess symptoms over time. 26 In line with other population-based studies and prior factor analyses, 11,13,27 a 3-item apathy subdimension has been used to screen for apathy symptoms, which included the following questions: (1) Have you dropped many of your activities and interests? (2) Do you prefer to stay at home rather than to go out and do new things? (3) Do you feel full of energy? (GDS-3A; range 0-3 points, with greater scores implying more apathy). Since scores ≥2 points have shown to be indicative of a clinical diagnosis of apathy, with a sensitivity of 69% and a specificity of 85%, 11,28 this cut-off has been applied to dichotomize the presence of apathy symptoms.
Likewise, the 12 remaining items of the GDS were summed up (i.e. those not reflecting apathy, the GDS-12D; range 0-12 points, with greater scores indicating more depression) and a ≥2 point cut-off was applied to reflect the presence of depressive symptoms.
As apathy is increasingly regarded as an independent syndrome across disease boundaries 2,9 and vascular apathy is preferably studied independently of depression, 12 we classified apathy symptoms without these depressive symptoms as isolated apathy (concurrent GDS-3A ≥ 2 and GDS-12D < 2). 8 Because the 15-item GDS is self-reported and its validity is decreased in older adults with dementia, 29 the GDS was either not administered to participants with a Mini-Mental State Examination (MMSE) score <19 points 11,30 or we excluded their GDS-scores to align data availability between the cohorts (Supplementary Table 1).
Exclusion of these GDS-scores was carried out by assigning all GDSmeasurements in which the concurrent MMSE-score was below 19 points to missing.

| Other measures
Data on demographic variables such as age, gender, educational attainment (up to and including secondary school vs. higher education) and living situation (institutionalized or not) was gathered during the baseline interviews. LiLACS NZ participants also reported their ethnicity, whether they identified themselves as Māori, which was prioritized over other ethnicities if more than one was listed. 21 According to overall data availability and previously applied conditions 11 participants were considered to have a chronic disease when any of the following diseases was present: arthritis, cancer, diabetes mellitus, chronic lung disease or Parkinson's disease. This was evaluated by combining reported histories, related medication use and blood glucose levels, which are specified in the supplement (Supplementary Table 1).

| Statistical analysis
A two-stage IPD approach was adopted to acknowledge the heterogeneity between the cohorts and to honour the preferences and practices in NZ Māori research, 31 which means the analyses were initially performed per cohort and their distinct results were afterwards pooled for meta-analysis.

| Cohort-level analyses
Generalized linear mixed models were applied as these mixed models optimize the use of available data during follow-up and account for dependency of repeated measures per person. 32 Two models were assessed in accordance with existing literature 11,13 : (1) adjusted for age (only in Māori), sex and depressive symptoms (GDS-12D ≥ 2, as a time-varying covariate) and employing a random intercept, and (2) additionally adjusted for educational attainment and chronic disease.
Besides the aforementioned exclusion of total GDS-scores based on MMSE-scores, no adjustment was made for cognitive status as it was thought to possibly be part of the causal pathway of interest. The selection of the final model is described in more detail in Supplementary Table 2. The cross-sectional (main effect of vascular disease) as well as the longitudinal estimates (interaction term: vascular disease * time since baseline) were generated by use of the Lme4 package in R (version 4.0.2). 33,34 These estimates were respectively presented as baseline and annual odds ratio (OR) of apathy symptoms (GDS-3A ≥ 2) with a 95% confidence interval (CI) (95% CI). The VAN DER KLEI ET AL.
-3 latter represents the additional OR with every year increase in time since baseline when having vascular disease and was stratified by baseline apathy status. All analyses were repeated while excluding participants with a history of stroke as a proxy for subsequent structural brain damage. 11,13

| Pooled analyses
The individual ORs were thereafter pooled by random-effects models with inverse-variance weighting and summarized in forest plots by using Review Manager 5.4. 35 Consistency between the estimates was defined as a statistical heterogeneity (I 2 ) below 40%. 36

| Sensitivity analyses
To examine whether our findings were dependent on the cut-offs that were applied to define apathy and depressive symptoms, we conducted sensitivity analyses by use of the ordinal package. 37 We entered the continuous instead of the dichotomized GDS-3A and GDS-12D scores in ordinal logistic regression models, which were apart from that equal to the main analyses.

| Demographic and clinical characteristics
Out of the 2576 participants of the TULIPS-consortium we excluded 706 older adults due to data unavailability (27.4% in total; exclusion on MMSE accounted for 1.5%). This meta-analysis therefore consisted of 1868 participants across the four cohorts and their baseline characteristics are shown in Table 1

| Sensitivity analyses
Sensitivity analyses applying the continuous GDS-3A and GDS-12D scores showed comparable results for both the cross-sectional and the longitudinal associations on vascular apathy (not shown).

| DISCUSSION
In this IPD meta-analysis of four population-based cohorts of the very old, we found that vascular disease is cross-sectionally associated with the presence of apathy symptoms, independent of concurrent depressive symptoms and only partially explained by a history of stroke. In contrast, vascular disease was not associated with the occurrence of apathy symptoms over time, which was also irrespective of whether apathy symptoms were experienced at baseline.
In relationship to previous transdiagnostic literature on the vascular apathy hypothesis, our significant cross-sectional associations are not only consistent in relating vascular disease to apathy in older populations, independent of depression and/or structural brain damage following stroke, but rather add to this body of literature by extending this finding to those of advanced age (i.e. in the very old vs. a mean age of 57 and 74 years old). 13,38 Our findings were also consistent in direction and magnitude throughout the four cohorts examined suggesting international relevance. This is surprising as the levels of apathetic and depressive symptomatology differed between the cohorts and performance of self-rated health tools can be culturally or socioeconomically mediated. 39  27 [24,28] 28 [26,29] 28 [27,29] 28 [26,29] 28 [26,29] Depressive symptoms, n Depressive symptoms were defined as a score of ≥2 on the GDS-12D, the depression subdimension of the GDS. d Apathy symptoms were defined as a score of ≥2 on the GDS-3A, the apathy subdimension of the GDS. e Isolated apathy was defined as having apathy symptoms in the absence of above depressive symptoms (i.e. GDS-3 A ≥ 2 and GDS-12D < 2). f Vascular disease was defined as a history of any clinical atherosclerotic pathology (including angina pectoris, myocardial infarction, intermittent claudication, transient ischemic attack, stroke and any related coronary or peripheral artery surgery). VAN DER KLEI ET AL.
-5 significant). This is not unexpected as NZ Māori have been exposed to a lifetime of discrimination resulting in health disparities, 41 but suggests nuance in its interpretation. As there are no other prospective, population-based studies to substantiate the vascular apathy hypothesis, 12 unequivocal understanding of a potential vascular ground for apathy symptoms remains limited. Optimizing CVD-management to attempt to prevent apathy in older adults may not be warranted, at least in advanced age.
There could be several pathophysiological mechanisms underlying our findings. One explanation is reversed causality, which has been evidently substantiated in the previously mentioned metaanalysis. 8 Apathy symptoms increased the risk of incident myocardial infarction and stroke by 21% and 37% respectively over a median follow-up of 8.8 years in approximately 48,000 older adults with a median age of 74 years old (IQR 72-78). To elucidate whether our cross-sectional results reflect this reversed causality or whether our lack of longitudinal evidence is, for instance, explained by characteristics of the TULIPS-consortium (such as competing risks at advanced age), future research could assess this reversed association for the very old and study vascular apathy in relatively younger, more stable populations of older adults over time (e.g. over 65 years old). In addition, apathy might be a marker of subclinical (cerebral) small vessel disease which could then result in overt CVD. 8,12 Support for this mechanism is strengthened by associations between more white matter hyperintensities on brain imaging and higher risk of having apathy, 12,42 which should also be studied prospectively in the general older population to elucidate the direction of their interplay. 12 Although we did not find support for vascular disease to predispose older adults to apathy symptoms, if the origin were to be vascular, mediation by cognitive decline could explain our lacking prospective associations as older people with severe cognitive impairment could not partake by design. To that extent, upcoming studies on apathy could be complemented by informant-rated scales or clinical interviews. Lastly, apathy could also be a marker or consequence of (multi)morbidity instead of specifically having a vascular origin; especially the apathy symptoms we studied, as they have been shown to closely connect to functional disability. 43 This calls for future apathy research to employ the diagnostic criteria 2 or instruments specifically designed to assess apathy in otherwise healthy, community-dwelling samples (e.g. the Apathy-Motivation Index) 44

F I G U R E 3
Longitudinal association between vascular disease at baseline and apathy symptoms during follow-up, stratified by baseline apathy status. Results from two-staged, generalized linear mixed model (GLMM) analyses, presented as annual odds ratios of apathy symptoms (GDS-3A ≥ 2) by vascular disease status (i.e. analyses including and excluding participants with a history of stroke). Fully adjusted models are shown (i.e. adjusted for depressive symptoms, age (only in Māori), sex, education and chronic diseases), stratified by baseline apathy status. Individual study estimates were pooled by random-effects models with inverse-variance weighting. a Baseline apathy status was defined as the presence or absence of apathy symptoms at baseline (i.e. a score of ≥2 or <2 on the GDS-3A, the apathy subdimension of the GDS). CI, confidence interval; GDS, Geriatric Depression Scale; LiLACS NZ, Life and Living in Advanced Age: a Cohort Study in New Zealand; OR, odds ratio. honoured by our two-stage IPD approach. Participants were solely excluded due to data unavailability and by using this IPD-method we verified previously presented results, 11 as well as provided consistency in the data and analyses across cohorts.
Although there was not much statistical heterogeneity between the pooled estimates, clinical heterogeneity between the cohorts is a potential limitation, which existed throughout time (i.e. initiation in different time periods) and at baseline (e.g. ethnicity and living situation). Furthermore, a disadvantage inherent to longitudinal research in the very old is the loss-to-follow-up due to death or ill health. 21 Subgroup analyses were hampered (e.g. minor vs. major vascular disease) and potential other confounders could not be included (e.g. social factors). The associations studied in the TULIPS-consortium could be even more unraveled by performing a systematic review to find other comparable cohorts. Another limitation is that our conclusions are confined to apathy symptoms of the three included questions in the GDS-3A and deficits in self-awareness, commonly associated with apathy, 7 could have impeded the reliability of this self-reported assessment. Despite these characteristics the GDS-3A has however been deemed appropriate for etiological research purposes, particularly in older populations over time. 11,45 In conclusion, vascular disease in the very old population is associated with apathy symptoms cross-sectionally, but there is no evidence of influence on apathy occurrence over time, independent of concomitant depressive symptoms. While these findings query the vascular apathy hypothesis and may discourage prevention of apathy by CVD-management, they emphasize the international relevance of their interplay in advanced age.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.