key: cord-0290405-xp413wcm authors: Shen, C.; Sahakian, B.; Cheng, W.; Kang, J.; Dong, G.; Xie, C.; Zhao, X.-M.; Feng, J. title: Social isolation, loneliness and all-cause dementia: a longitudinal and imaging-genetic study in the UK Biobank cohort date: 2021-07-06 journal: nan DOI: 10.1101/2021.06.30.21259818 sha: 0eb53b49a195f32d1eda6dd2d7c82894beae3698 doc_id: 290405 cord_uid: xp413wcm INTRODUCTION: Current findings of the relative influence of social isolation and loneliness on dementia are contradictory, and the potential neurobiological mechanisms are unclear. METHODS: We utilized the UK Biobank to investigate the relationships of social isolation and loneliness with dementia (n = 462,619). Neuroanatomical correlates were identified in a subset of participants (n = 32,263). The transcriptomic signatures of related brain changes were characterized by gene enrichment analysis. RESULTS: After full adjustment, social isolation but not loneliness was associated with dementia (hazard ratio: 1.28, 95% confidence interval: 1.17-1.39). Isolated individuals had reduced gray matter volumes in temporal, frontal, occipital and subcortical regions (e.g., hippocampus and amygdala). Relevant brain changes were spatially correlated with genes involved in mitochondrial dysfunction and oxidative phosphorylation, and down-regulated Alzheimer's disease-related genes. DISCUSSION: Social isolation is an independent risk factor for dementia, which could be partly explained by related structural changes coupling with altered molecular functions. Social isolation (objective reflection of social relationships) and loneliness (perceived social isolation) are serious yet underappreciated public health problems that are particularly associated with old age. 1 Dementia is a major cause of disability in the elderly, affecting over 46 million people worldwide in 2015 and was estimated to increase to 131.5 million by 2050. 2 To date, the influence of social isolation and loneliness on dementia is still unclear. Whereas in some studies social isolation but not loneliness was associated with increased risk of dementia and cognitive decline, 3, 4 other studies found the opposite result. 5 One possibility for this discrepancy is that the associations may be impacted by risk factors which were not consistently considered in past studies. To what extent established risk factors account for such associations requires further elucidation. Moreover, little is known about the underlying neurobiological mechanisms. The social brain hypothesis posits that the evolution of human brain is driven by increasingly complex social selection pressures. 6 Thus, social isolation may impair the brain regions involved in higher cognitive and affective processes, and in turn affect cognitive abilities. Only a paucity of studies have explored the neural underpinnings of social isolation and loneliness. Structural and functional changes in several brain regions including prefrontal, temporal and parietal cortices, amygdala, hippocampus, striatum and ventral tegmental area were reported. [7] [8] [9] [10] [11] [12] Spreng et al. (2020) 13 reported that loneliness is associated with the default network. However, findings are mixed probably due to the heterogeneity of methods and study design. 14 Social factors may also play a significant role in regulating the transcriptional activity of human genome. 15 Previous gene expression studies in post-mortem nucleus accumbens and All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.06.30.21259818 doi: medRxiv preprint 5 dorsolateral prefrontal cortex revealed that loneliness-related differentially expressed genes were associated with Alzheimer's disease (AD) and immune dysfunction. 16, 17 One recent approach leveraged brain-wide gene expression atlases to link molecular function to macroscale brain organization, 18 which was helpful to understand disease-related brain alterations and has been used in depression 19 and neurodegenerative diseases. 20 The present study utilized the UK Biobank cohort to examine the relative influence of social isolation and loneliness on incident dementia, and to quantity the extent to which these associations were explained by various risk factors such as biological and psychological factors. Next, we investigated the neuroanatomical correlates of social isolation and loneliness, and tested whether the identified brain alterations could mediate the prospective associations of social isolation and loneliness with cognitive abilities. Finally, using the Allen Human Brain Atlas (AHBA) microarray dataset, 21 we analyzed the gene expressions of which biological processes or functional pathways were associated with social isolation-and loneliness-related brain changes. Besides, we specifically examined the associations between brain changes and AD-related genes. The UK Biobank is a prospective epidemiological study that involves over 500,000 individuals recruited in 22 centers across the UK between 2006 and 2010. 22 The study has collected extensive questionnaire data, physical measurements, and biological samples. A subset of the cohort has been invited back to collect multimodal imaging data and repeat behavioral assessments since 2014. All participants are followed up for health conditions through linkage to All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.06.30.21259818 doi: medRxiv preprint 6 national electronic health-related datasets. The health follow-up data used in the present study started at enrollment and ended in January 2021. All participants provided informed consents and the ethical approval was from the North West Multi-Centre Research Ethics Committee. Social isolation and loneliness were calculated from scales that were used in previous studies. 23 Social isolation was assessed with three questions: number of people living together in the household (1 = living alone), frequency of visiting friends or family or having they visit you (1 = less than once a month) and attending leisure or social activities at least weekly (1 = no participation). An individual was defined as socially isolated if he or she scored 2 or 3, and those who scored 0 or 1 were classified as not isolated. Loneliness was constructed from two questions: often feeling lonely (1 = yes) and frequency of confiding in close people (1 = less than once every few months). An individual was defined as lonely if he or she scored 2, and not lonely if he or she scored 0 or 1. All-cause dementia was identified as International Classification of Diseases 10th codes F00, F01, F02, F03 or G30 of the "first occurrence" data-fields generated by the UK Biobank (data version: January 2021; https://biobank.ndph.ox.ac.uk/showcase/label.cgi?id=1712), which were ascertained by the combination of primary care, hospital in-patient, death register and self-reported data. Date of diagnosis was set as the earliest date of dementia codes recorded regardless of source used. Prevalent dementia cases were defined as the date of diagnosis occurred All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.06.30.21259818 doi: medRxiv preprint 7 within the first 3 years of follow-up or self-reported "dementia, AD or cognitive impairment" at baseline, and were excluded to avoid the possible reverse-causation bias (n = 494). The source of incident all-cause dementia cases was reported in Table S1 . Pairs matching and reaction time tests were administered to almost all participants at baseline and over 40,000 participants at the imaging visit. In the first test, respondents were asked to correctly identify matches from six pairs of cards after they had memorized their positions. The number of incorrect matches was recorded, with a greater number reflective of a poorer visual memory. This visual memory error score was logarithmically transformed (In(x+1)) due to skewed distribution and zero-inflation. The card-game "Snap" asked participants to press a button when two simultaneously presented cards matched, with shorter reaction time represented faster speed of processing. The log-transformed mean reaction time over 12 rounds was used (In(x)). In addition, these two tests were normalized and then averaged to yield a composite score of cognition. Neuroimaging data were collected across three dedicated, identical imaging centers. The detailed MRI acquisition protocol has been described elsewhere. 24, 25 All structural MRI data were preprocessed in the Statistical Parametric Mapping software version 12 using the CAT12 toolbox with default settings, including the usage of high-dimensional spatial normalization with an already integrated Dartel template in Montreal Neurological Institute (MNI) space. All images All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.06.30.21259818 doi: medRxiv preprint 8 were subjected to nonlinear modulations and corrected for each individual head size. Images were then smoothed with an 8mm full-width at half-maximum Gaussian kernel with the resulting voxel size of 1.5mm 3 . We focused our analyses within the automated anatomical labelling atlas 3 26 excluding cerebellum. We used the transcriptomic data from six neurotypical adult brains in the AHBA. 21 Tissue samples in left hemisphere were used due to the right hemisphere data were only available for two donors. The preprocessing steps included probe-to-gene re-annotation, intensity-based data filtering, probe selection by mean, separating tissue samples into subcortical and cortical regions, and within-donor normalization, as reported previously. 27 Finally, we obtained expression values for 15,408 unique genes at 711 cortical and 135 subcortical locations separately. Associations of social isolation and loneliness with dementia incidence were investigated using Cox proportional hazard models with age as a timescale. The results were presented as hazard ratios (HRs) and 95% confidence intervals (CI). The proportional hazard assumptions were checked using Schoenfeld residuals, and no major violations were observed. We restricted analyses to participants who had complete data on baseline social isolation, loneliness and cognitive tests, and incident dementia. The minimal model was adjusted for age, sex and ethnicity. To assess the extent to which other risk factors explained the associations, percentage of excess All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.06.30.21259818 doi: medRxiv preprint 9 risk mediated (PERM) was calculated for: (1) socioeconomic factors (education and household income); (2) biological factors (BMI, blood pressure, chronic illness and APOE genotype); (3) cognitive factors; (4) health behavior (alcohol intake, current smoker and physical activity); (5) depressive symptoms; (6) loneliness (for social isolation as the exposure) or social isolation (for loneliness as the exposure). Covariates were treated as categorical variables with missing value as a separate category (Method S1 and Table S2 ). Finally, a full model including all risk factors was conducted. PERM was estimated using the formula: [HR (age, sex and ethnicity adjusted) -HR (age, sex, ethnicity and risk factor adjusted) ]/[HR (age, sex and ethnicity adjusted) -1] 100. 28 We also performed subgroup analyses to assess potential modification effects by sex (men or women) and age (< 60 or 60 years), and a sensitivity analysis using complete cases to test the robustness. Calculations were performed by R version 3.6.0 using the survival package. 29 Linear regression models were conducted to investigate the cross-sectional associations of social isolation and loneliness with gray matter volume (GMV) separately. Age, education and household income at the imaging visit, and sex, ethnicity, site and total intracranial volume (TIV) were used as covariates of no interest. Multiple comparison correction was performed at voxel level with a P FDR < .05. A significant cluster was defined on the basis of an 18-connectivity criterion 30 and having more than 217 voxels falling into the 90% CI of the smoothing kernel voxels. 27 Using the data-driven approach, we found significant GMVs associated with social isolation but no regions for loneliness. Three sensitivity analyses were performed: (1) excluding participants with dementia at any time (n = 29), (2) additionally adjusted for loneliness, (3) All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.06.30.21259818 doi: medRxiv preprint additionally adjusted for depressive symptoms. The identified brain regions were decoded for cognitive functions using Neurosynth meta-analysis toolbox. 31 We calculated cross-voxel Spearman's correlations between the significant social isolation t-statistic map and the association Z-statistic maps. The top 100 negative terms were selected, from which the anatomical terms such as "hippocampus" were removed, and the retained cognitive terms were visualized on a word-cloud plot (Method S2). Additionally, we tested the cross-sectional relationship between the significant GMVs and cognitive performance in the UK Biobank controlling for the same covariates as in the whole-brain analysis. Finally, the mediation effect of the significant GMVs on the association between social isolation at baseline and cognitive performance at the imaging visit was examined. Age, education and household income at baseline, time interval between baseline assessment and imaging collection, and sex, ethnicity, site and TIV were adjusted for. The significance of the mediation was estimated by 10,000 bias-corrected bootstrapping, which was performed using the mediation toolbox developed by Wager et al.. 32 Partial least square (PLS) regression was used to relate the social isolation t-statistic map (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.06.30.21259818 doi: medRxiv preprint 11 PLS using 5000 null t-statistic maps which were obtained by label shuffling for social isolation, and counting the number of times the explained variance was higher than the original observation (denoted as P perm ). 33 As PLS1 was not significant in subcortical regions, the following analyses were limited in cortical regions. Bootstrapping (5000 times) was used to estimate the variability of each gene's PLS1 weight, and the ratio of the weight to its bootstrapped standard error was used to calculate the Z-score. 34 Genes with PLS1 weights Z > 4 (PLS1+) or Z < -4 (PLS1-) (all P FDR < .001) were used to calculate enrichments in both Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways and Gene Ontology (GO) terms of biological processes (Method S3). FDR correction was performed for KEGG pathways and GO biological processes simultaneously. Significance was set at P FDR < .05. Specially, we examined whether AD-related dysregulated genes were expressed most in regions that were morphometrically correlated to social isolation. A recent systematic integrated analysis reported 2444 up-regulated and 2978 down-regulated AD-related genes (Method S4). 35 Furthermore, we investigated the relationship of social isolation-related GMV changes with the average expression level of up-and down-regulated AD-related genes separately. The significance was tested by 5000 times permutation, in which the null distribution was defined by 5000 random t-statistic maps described above. At baseline, 462,619 participants provided complete data on social isolation, loneliness, cognitive tests and incident all-cause dementia ( Table 1 and Table S3 ). Of these, 41,886 (9%) All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. (Table S4 ). At the imaging visit, 2371 (7%) participants were socially isolated and 1503 (5%) participants were lonely. After adjustment for age, sex and ethnicity, the HR for incident all-cause dementia was 1.64 (95% CI: 1.51-1.78) for social isolation compared with no social isolation. Further adjustment for different risk factors attenuated the association (PERM: 8%-31%). The overall attenuation after adjustment for all risk factors was 56% to 1.28 (95% CI: 1.17-1.39; Figure 1 ). In contrast, the minimally adjusted HR for loneliness was 1.55 (95% CI: 1.40-1.71), but reduced to nonsignificance when all risk factors were included (P = .45). Adjustment for depressive symptoms had the greatest impact on the association, which was decreased by 75% ( Figure 1 ). Subgroup analyses found the association between social isolation and dementia was consistent across sex ( Figure S1 ). However, the association was only significant in the elderly group with full adjustment (HR: 1.30, 95% CI: 1.18-1.42; Figure S2 ). Consistent findings were found in the complete case analyses ( Figure S3 ). All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The whole-brain analysis revealed that brain structures encompassing temporal, frontal, occipital and subcortical regions were associated with social isolation (Figure 2A and Table S5 ). The most significant regions were located in bilateral hippocampus and parahippocampus, left thalamus, fusiform and amygdala, and right supramarginal gyrus. The findings were robust if excluding patients with dementia ( Figure S4) , and similar brain regions were found if further adjusted for loneliness ( Figure S5 ) or depressive symptoms ( Figure S6 ). Using Neurosynth, we found the significant regions where GMV was smaller in socially isolated individuals than controls tend to be involved in memory and learning tasks ( Figure 2B and Table S6 ). In the UK Biobank, the total GMV of the significant brain regions was associated with cognitive performance at the imaging visit (r p (30469) = -0.04, P < .001; Figure 2C ), and could significantly mediate the prospective relationship between social isolation and cognitive performance (path ab: β = 0.003, 4% of the total effect size, P < .001; Figure 3 ). In cortical regions, the PLS1 was significant (i.e., explained 15% of the variance of social isolation-related structural changes, P perm < .05; Result S1). Based on the normalized PLS1 weights, there were 2029 genes in the PLS1+ gene set (Z > 4) and 1048 genes in the PLS1-gene set (Z < -4) (all P FDR < .001; Figure S7 and Table S7 ). After the FDR correction, the PLS1+ gene set was enriched in 38 biological processes such as "mitochondrion organization" and "oxidative phosphorylation", and 6 KEGG pathways such as "Parkinson's disease" and "Alzheimer's disease" ( Figure 4A and Table S8 ). No significant enrichment results for the PLS1-gene set. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.06.30.21259818 doi: medRxiv preprint Moreover, PLS1+ gene set was highly enriched among genes that were recently reported as underexpressed in post mortem brain tissue from patients with AD (P FDR < .001; Table S9 ). Social isolation-related changes in regional GMV was positively associated with cortical gene expression of down-regulated AD-related genes (r s (711) = 0.13, P perm = .02; Figure 4B ). Nevertheless, no significant spatial correlation with up-regulated AD-related genes was found (P perm = .28; Figure 4C ). This study suggested that social isolation but not loneliness was associated with elevated risk of incident all-cause dementia. Socially isolated individuals presented reduced GMVs of brain regions involved in memory and learning, which partly mediated the prospective relationship between social isolation and cognitive performance. Transcriptomic analyses found that social isolation-related brain changes were spatially correlated with genes associated with mitochondrial function and oxidative stress, and with AD-related down-regulated genes. Taken together, evidence from behavior, neuroimaging and transcriptome revealed that social isolation was an independent risk factor of cognitive decline and dementia. In the context of COVID-19 pandemic which critically exacerbated social isolation, our findings may provide insights into effective intervention strategies especially for old people. To our knowledge, this is the first large-scale study clarifying the contribution of established risk factors to associations of social isolation and loneliness with dementia. Consistent with previous longitudinal studies, 36 , 37 we found social isolation was associated with 1.28-fold increased risk of developing dementia, which was independent of loneliness and other risk factors. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.06.30.21259818 doi: medRxiv preprint However, loneliness was not related to incident dementia after full adjustment, owing to 75% of the relationship was attributable to depressive symptoms. Similar results were reported in studies using mortality as the outcome. 23, 38 Social isolation and loneliness were often weakly correlated, 39 our results supported that these two might be independent constructs. We identified multiple brain regions associated with social isolation in the largest sample to date, while no results for loneliness, implying that objective and subjective social relationships possibly have distinct neural bases. In the UK Biobank, socially isolated individuals had the most severe gray matter atrophy in hippocampus, parahippocampus, thalamus, amygdala and temporal cortex, which was not confounded by loneliness and depressive symptoms. The impact of social stressor on hippocampal morphology has been well documented. 40, 41 Amygdala played an essential role in emotional processing, and was relevant to social network size. 9, 10 Temporal cortex such as fusiform and superior temporal gyrus was important for social perception. 42, 43 Experimental manipulation of group size in macaques resulted in variation in the volume of mid-superior temporal sulcus and amygdala. 44 Cognitive annotation of the significant brain map demonstrated that impaired regions related to social isolation were involved in cognitive processes especially memory and learning. Indeed, we found the volume of these regions was positively associated with cognitive performance, and could partly mediated the prospective association between social isolation at baseline and cognitive performance at the follow-up. Past epidemiological studies have revealed that social isolation was associated with cognitive function in later life, 45 but evidence on the potential mechanism is sparse. Animal studies suggested that isolation affected cognition via altering the excitatory and inhibitory synaptic density in hippocampus, 46 and social interaction rescued All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.06.30.21259818 doi: medRxiv preprint 16 memory deficit by increasing hippocampal neurogenesis. 47 The brain reserve hypothesis of dementia posits that larger brains have more neural matter that may increase tolerance of pathology and maintain cognitive function. 48 Our findings implicated that isolation, as a social stressor, might reduce available brain reserve to accelerate cognitive decline, and finally lead to increased risk of dementia. The gene expression profile of the social isolation-related cortical changes was enriched in biological processes and KEGG pathways closely linked to dementia. It is well established that mitochondrial dysfunction and oxidative damage is critical in aging and neurodegenerative diseases. 49 In AD, dysfunctional mitochondria released oxidizing free radicals in the brain and caused considerable oxidative stress. 50 Animal research showed that impairments of mitochondrial function 51 and oxidative phosphorylation 52 preceded the development of AD pathology. We found more down-regulated AD-related genes in the list of positively weighted genes with changes in cortical GMVs. And the expression level of down-regulated genes was significantly associated with cortical changes. These results suggested that genes with reduced brain postmortem transcription in AD were underexpressed in cortical regions with higher levels of changes related to social isolation. Notably, the pattern of brain alterations was obtained from population-based samples, indicating that the brain changes associated with social isolation might precede the diagnosis of dementia. Some limitations should be mentioned. It was not able to discriminate acute and chronic social isolation in the UK Biobank. Evolutionary theory believes that acute isolation will induce adaptive physiological responses such as increased inflammation and hypothalamic-pituitary-adrenal axis activation to motivate individuals to restore social All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.06.30.21259818 doi: medRxiv preprint connections. 53 While chronic isolation perhaps leads to the set-point adaptations due to the long-term failure to maintain social homeostasis. 54 The temporal dynamics of social isolation may link to different neural activity and health consequences. Akhter-Khan et al. 55 reported that transient loneliness was resilient to dementia risk. Consequently, future research should consider the trajectory of social isolation and loneliness. Older adults are particularly vulnerable to COVID-19, and therefore subject to greater social restrictions. It is hypothesized that the virus may increase the risk of cognitive decline and dementia by affecting the brain. 56 However, the corresponding social isolation measures may also impact the risk of atypical aging and dementia in normal people. Although we currently do not have sufficient data to answer this question, continued monitoring is necessary to investigate the relationship between COVID-19 and dementia. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.06.30.21259818 doi: medRxiv preprint Figure 3 Mediation analysis using the identified gray matter volume as a mediator of the prospective relationship between social isolation and cognitive performance (n=30,612). The predictor in the model is social isolation at baseline, the mediator is the total of significant GMVs associated with social isolation at the imaging visit, and the dependent variable is cognitive performance at the imaging visit, which was a composite score of visual memory and reaction time. Covariates included age, education and household income at baseline, time interval between the baseline assessment and imaging collection, and sex, ethnicity, site and TIV. Path a measures the association between the predictor and the mediator; path b represents the effect of the mediator on the dependent variable while controlling for the predictor; path c measures the total relationship between the predictor and the dependent variable; path c' measures the direct effect; the mediation effect is the product of path a and path b (a*b). SI = social isolation. **P < .01, ***P < .001. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.06.30.21259818 doi: medRxiv preprint Association between social isolation t-value and the average expression level of up-regulated AD-related genes in cortical regions. The significance was tested by 5000 times permutation, in which the null distribution was defined by 5000 random t-statistic maps obtained by label shuffling for social isolation. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.06.30.21259818 doi: medRxiv preprint Social isolation and loneliness in older adults: Opportunities for the health care system The global impact of dementia: an analysis of prevalence, incidence, cost and trends The association between social engagement, loneliness, and risk of dementia: a systematic review and meta-analysis Social isolation, rather than loneliness, is associated with cognitive decline in older adults: the China Health and Retirement Longitudinal Study Feelings of loneliness, but not social isolation, predict dementia onset: results from the Amsterdam Study of the Elderly (AMSTEL) Evolution in the social brain Structural brain correlates of loneliness among older adults Online social network size is reflected in human brain structure Long-term effects of social stress on brain and behavior: a focus on hippocampal functioning Brain Changes in Response to Long Antarctic Expeditions Stress and the social brain: behavioural effects and neurobiological mechanisms Social perception from visual cues: role of the STS region Social network size affects neural circuits in macaques Social isolation and cognitive function in later life: a systematic review and meta-analysis Enriched environment and social isolation affect cognition ability via altering excitatory and inhibitory synaptic density in mice hippocampus Social interaction rescues memory deficit in an animal model of Alzheimer's disease by increasing BDNF-dependent hippocampal neurogenesis All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted Mitochondria take center stage in aging and neurodegeneration Alzheimer's disease Mitochondrial bioenergetic deficit precedes Alzheimer's pathology in female mouse model of Alzheimer's disease Oxidative damage is the earliest event in Alzheimer disease Loneliness in the modern age: an evolutionary theory of loneliness The neural circuitry of social homeostasis: Consequences of acute versus chronic social isolation Associations of loneliness with risk of Alzheimer's disease dementia in the Framingham Heart Study All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted 6%) 11,896 (28.4%) With chronic illness All rights reserved. No reuse allowed without permission.(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.The copyright holder for this preprint this version posted July 6, 2021. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.The copyright holder for this preprint this version posted July 6, 2021. All rights reserved. No reuse allowed without permission.(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.The copyright holder for this preprint this version posted July 6, 2021. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.The copyright holder for this preprint this version posted July 6, 2021. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.The copyright holder for this preprint this version posted July 6, 2021. All rights reserved. No reuse allowed without permission.(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.The copyright holder for this preprint this version posted July 6, 2021. HR=hazard ratio. PERM=percentage of excess risk mediated. *: adjusted for age, sex and ethnicity.All rights reserved. No reuse allowed without permission.(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.06.30.21259818 doi: medRxiv preprint 28 Figure 2 Neuroanatomical correlates of social isolation at the imaging visit (n=32,263). A. Results of whole-brain voxel-wise analysis of social isolation. Age, education and household income at the imaging visit, and sex, ethnicity, site and TIV were used as covariates. Multiple comparison correction was performed at voxel level with a P FDR < .05, and a cluster was significant with more than 217 voxels. Significant GMVs in cortical and subcortical regions are presented separately.The color bar presents t-value. B. Word-cloud plot of cognitive terms associated with social isolation related GMVs. The size of cognitive terms corresponds to the Spearmen's correlation of corresponding meta-analytic maps in Neurosynth with the significant social isolation t-statistic map. C. Cross-sectional association between social isolation related GMVs and cognitive performance in the UK Biobank (n=30,480). Cognitive performance was a composite score of visual memory and reaction time. The partial correlation coefficient with the same covariates as in All rights reserved. No reuse allowed without permission.(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.06.30.21259818 doi: medRxiv preprint 29 the whole-brain analysis is showed. The color bar presents the estimated probability density.All rights reserved. No reuse allowed without permission.(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.The copyright holder for this preprint this version posted July 6, 2021. ; https://doi.org/10.1101/2021.06.30.21259818 doi: medRxiv preprint