key: cord-0425959-d8wna2s1 authors: Lammer, L.; Beyer, F.; Luppa, M.; Sander, C.; Baber, R.; Engel, C.; Wirkner, K.; Loeffler, M.; Riedel-Heller, S.; Villringer, A.; Witte, V. title: Social isolation and the aging brain. Social isolation is linked to declining grey matter structure and cognitive functions in the LIFE-Adult panel study date: 2021-12-16 journal: nan DOI: 10.1101/2021.12.14.21267787 sha: 3e705928085f3f49dc86b2d48a1d7241d8f46f52 doc_id: 425959 cord_uid: d8wna2s1 Background Social isolation is a risk factor for dementia, a devastating disease with a rapidly growing global prevalence. However, the link between social isolation and changes in brain structure and function is poorly understood, as studies are scarce in number, methodologically inconsistent and small in size. In this pre-registered analysis of a large population-based panel study, we aimed to determine the impact of social isolation on brain structures and cognitive functions central to age associated decline and dementia. Methods and findings We analysed data of 1992 cognitively healthy participants of the LIFE-Adult study at baseline (age range: 50-82 years) and of 1409 particpants at follow-up (average change in age: 5.89 years). We measured social isolation using the 30-point Lubben Social Network Scale (LSNS) and derived measures of grey matter structure from anatomical 3T MRIs. We employed covariate adjusted linear mixed models to test the associations of baseline social isolation and change in social isolation with hippocampal volume, cognitive functions (executive functions, memory, processing speed) and cortical thickness. We found stronger baseline social isolation to be significantly associated with smaller hippocampal volumes ({beta} = -5.5 mm3/LSNS point(pt), FDR q = 0.004, BF = 14.6) and lower cognitive functions (all {beta} < -0.014 SD/pt, FDR q < 0.003, BF > 49). Increases in social isolation over time were linked to hippocampal volume decline ({beta} = -4.9 mm3/pt, FDR q = 0.01, BF = 2.9) and worse memory performance ({beta} = -0.013 SD/pt, FDR q = 0.04, BF = 1.1). Furthermore, we detected a significant interaction of baseline social isolation with change in age on hippocampal volume ({beta} = -0.556 mm3/pt*a, q = 0.04, BF = 0.5), indicating accelerated brain aging in more isolated individuals. Moreover, social isolation cross-sectionally and longitudinally correlated with lower cortical thickness in multiple clusters in the orbitofrontal cortex, precuneus and other areas (FDR q < 0.05). Conclusions Here, we provide evidence that social isolation contributes to hippocampal and cortical atrophy and subtle cognitive decline in non-demented mid- to late-life adults. Importantly, within-subject effects of social isolation were similar to between-subject effects, indicating an opportunity for targeting social isolation to reduce dementia risk. Over 50 million humans suffer from dementia today. In just 20 years this number will likely double. Already now, dementia's global annual costs exceed one trillion US$(1) and its detrimental effects on the lives of the afflicted makes it a major contributor to the world's burden of disease (2) . Research on pharmacological interventions targeting dementia pathogenesis have not yielded any result with a clear clinical benefit yet (3, 4) and available drugs targeting cognitive symptoms offer at most a minor alleviation (5, 6) . Henceforth, prevention is of cardinal importance and potentially modifiable risk factors are our most promising target (7) . Systematic reviews and meta-analyses have concluded that social isolation, the objective lack of social contact, is such a risk factor for dementia (8, 9) and its main feature cognitive decline (10) (11) (12) . Assuming causal relationships, Livingston et al. calculated population attributable fractions for risk factors for dementia and concluded that 3.5% of cases could be attributed to social isolation. This is almost as many as to obesity, hypertension and diabetes combined. Dementia is characterized by progressing cerebral alterations such as vascular degeneration, amyloid plaques, tau fibrillary tangles and neural degeneration and grey matter loss. Neuroimaging correlates of these brain changes have been observed multiple years prior to symptom onset in autosomal dominant dementia (13, 14) and can already be detected in cognitively healthy persons using neuroimaging (15) . Thus, brain magnetic resonance imaging (MRI) can be a potent dementia-risk indicator (16, 17) , might offer pivotal guidance to identify patients for intensive dementia prevention (18) and serve as secondary outcome for intervention trials (19, 20) . However, as the US National Academy of Sciences stresses, this application requires further study(21). Still, the link between brain structure and social connections, the umbrella term encompassing social isolation, social support and loneliness, has not received much attention(22). First cross-sectional studies have linked low social connection to an elevated "brain age" gap estimate(23), changes in microstructural (24) (25) (26) and volumetric measures in brain regions including the hippocampus and the prefrontal cortex (26) (27) (28) (29) (30) (31) (32) . In a longitudinal study microstructural deteriorations and a larger total white matter hyperintensity volume were correlated with decreases in predominantly social activities. Furthermore, it suggested that white matter changes mediated the positive association between social activities and . CC-BY-NC 4.0 International license It is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted December 16, 2021. ; https://doi.org/10.1101/2021.12.14.21267787 doi: medRxiv preprint perceptual speed in this sample of 70 participants (37 at follow-up) > 80 years old (33) . Mortimer et al. conducted a small RCT with older adults and found increased total brain volumes and cognitive function in participants after a social interaction intervention compared to a non-intervention control group (34) . Taken together, the current evidence does not allow us to draw any firm conclusions due to various shortcomings. Researchers investigated a multitude of different outcome measures, but replication studies confirming previous results are still lacking despite their importance to neuroscience (35) . Moreover, the evidential value is often limited as most of the existing studies on social connection and brain structure are cross-sectional with relatively small sample sizes. The few exceptions to this do not investigate social isolation but rather related but distinct concepts like social activities. Moreover, no solid evidence on the mechanistic underpinnings of this relationship exists. Several mutually non-exclusive, partly overlapping theories are used to explain the beneficial effects of social interaction. The main-effect theory postulates that social relationships foster beneficial health behaviours, affective states and neuroendocrine responses, ultimately protecting neuronal tissue (36) . Others point out that socializing is cognitively demanding and requires engagement with complex environments. In the "use-it-or-lose-it" theory, this is crucial for the maintenance of cognitive function (37) . The stress-buffering hypothesis puts forward the beneficial effects of social support in strenuous times on mental, cognitive, and immunological health (36) . Lastly, reverse causality (38) or simultaneity (39) can underly the observed associations, too. This is sometimes referred to as health selection (9) in analogy to the homonymous economic concept (40) . In sum, reliable evidence based on longitudinal studies is required to assess effect sizes and disentangle correlation from causation to better understand the impact of social isolation on brain and cognitive aging. Thus, in this pre-registered analysis, we aimed to determine the relationship between social isolation, brain structure and cognitive functions in a large wellcharacterized longitudinal population-based sample of healthy mid-to late-life individuals. To this end, we investigated the connection of social isolation and hippocampal volume, a focal points of age-related atrophy and Alzheimer's disease (41) , cognitive functions and cortical thickness while adjusting for control variables including baseline age, change in age and gender. We hypothesized that both baseline and change in social isolation would be correlated with reduced hippocampal volume, cognitive functions (memory, processing speed, executive functions) and cortical thickness. Additionally, we hypothesized interaction effects of baseline social isolation with change in age in the same direction. Moreover, we aimed to test a mediating role of chronic stress as well as hippocampal volume in these models. We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) (42) and Committee on Best Practices in Data Analysis and Sharing (COBIDAS)(43) on MRI guidelines in our reporting wherever appropriately applicable. The study's preregistration can be found on https://osf.io/8h5v3/. Please refer to it for information on the authors' previous knowledge of the data and a comprehensive overview of our pre-specified hypotheses and models. We used longitudinal data from the "Health Study of the Leipzig Research Centre for Civilization Diseases" (LIFE). The LIFE-Adult-Study is a population-based panel study of around 10,000 randomly selected participants from Leipzig, a major city with 550,000 inhabitants in Germany. A sub-group of around 2600 participants underwent MRI testing at baseline. The baseline examination was conducted from August 2011 to November 2014. Follow-up assessments were performed around six to seven years after the respective first examinations (44) . Around 1000 participants of the MRI-subsample returned for follow-up testing. As changes in hippocampal volumes, our key region of interest, show accelerated reductions starting at about 50 years of age (45) , we included all participants over 50 with MRI data that did not fulfil any of the following exclusion criteria: -Anamnestic history of stroke any medical condition (i.e., epilepsy, Multiple sclerosis, Parkinson's disease) / chronic medication use that would compromise cognitive testing -(i.e., cancer treatment in the past twelve months or drugs affecting the central nervous system) diagnosed dementia or Mini-Mental State Examination (MMSE)-score <24 a trained radiologist considered the MRI scans unusable due to brain tumors, or acute ischemic, hemorrhagic or traumatic lesions . CC-BY-NC 4.0 International license It is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint If no MMSE data was available, the participants were excluded if their overall performance in cognitive tests negatively deviated from the wave's mean by 2 standard deviations (SDs). The exclusion criteria were chosen to reduce the potential of reverse causality as correlations observed in this cognitively healthy sample should not stem from prodromal dementia. We obtained T1-weighted images on a 3 Tesla Siemens Verio MRI scanner (Siemens Healthcare, Erlangen, Germany) with a 3D MPRAGE protocol and the following parameters: FreeSurfer's longitudinal pipeline on all scans, including those of participants without a follow-up scan (46) . Moreover, we smoothed the cortical thickness surfaces with a 10mm kernel to improve reliability and power (47) . Different Linux kernels and Ubuntu versions constituted the computational infrastructure during the data acquisition and processing. Visual quality control was based on the recommendations of Klapwijk et al. (48) . After the baseline data were acquired, our team visually controlled all results of the cross-sectional recon-all pipeline. Additionally, we controlled the outputs of the longitudinal stream of all participants with follow-up data and those whose cross-sectional runs required editing. If we detected errors in the processed scans, we manually edited them (N=283). We excluded participants from analyses using MRI measures if we deemed the processed scans to be unusable (n=68). We used the standard Lubben Social Network Scale (LSNS) -6(49) to measure the participants' social isolation. The questionnaire is a suitable tool to measure social isolation (50) and has a maximum score of 30. In order to make larger scores imply more isolation, we subtracted the actual score from the maximum score of 30. To quantify changes in social isolation, we subtracted the baseline from the follow-up score. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted December 16, 2021. ; https://doi.org/10.1101/2021.12.14.21267787 doi: medRxiv preprint We used the hippocampal volume derived from FreeSurfer's segmentation and averaged it over both hemispheres. Furthermore, we adjusted it for intracranial volume according to the following formula: where β is the unstandardized regression coefficient of hippocampal volume (HCV) on intracranial volume (ICV) from a linear mixed-effects model (LME) (51) . For whole brain analyses we used the FreeSurfer fsaverage template and cortical thickness as a vertex-wise outcome. The cognitive assessment consisted of the trail-making-test (TMT)(52) and the CERAD -plus test-battery (CERAD -Consortium to Establish a Registry for Alzheimer's Disease, RRID:SCR_003016). We z-transformed the results of the individual tests using the mean and SD derived from all measures. We calculated domain-specific summary scores and calculated them as follows (53, 54) : For the memory score, we defined learning as the sum of three consecutive learning trials of the CERAD word list (10 words), recall as the sum of correctly recalled words after a delay, in which participants performed a nonverbal task, and recognition as the number of correctly recognized words out of a list of 20 presented afterwards. memory = (z_learning + z_recall + z_recognition)/3 Processing speed was defined as the negated z-scored TMT part A score. As mentioned above, we also calculated a summary score to exclude cognitively impaired participants: Sum-score = z_phonemic fluency + z_semantic fluency + z_sum_learning + z_recall + . CC-BY-NC 4.0 International license It is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted December 16, 2021. ; https://doi.org/10.1101/2021.12.14.21267787 doi: medRxiv preprint Trierer Inventar zum chronischen Stress (TICS) is a German questionnaire assessing perceived stress (57 items, 0-4 points per item) (55) . Its sum score is our measure of participants' stress. Month and year of birth of the participants and the date of the MRIs were recorded and used For follow-up observations, we calculated the change in age as baseline age subtracted from follow-up age. For all baseline observation change in age = 0. Data on the following variables was only available for the baseline. Henceforth, we used the baseline values of these control variables for both timepoints. We calculated the body-mass-index (BMI) according to the standard formula: BMI = weight In order to control for hypertension and diabetes, we used dichotomized variables. Participants were categorised as hypertensive if they had a previous diagnosis of hypertension, took antihypertensive medication or had an average systolic blood pressure over 160mmHg. The systolic blood pressure was measured three times. The first measurement was performed after 5 minutes of rest and 3 additional minutes of rest passed between each of the following measurements. Participants were categorised as diabetic if they had a previous diagnosis of diabetes, took antidiabetic medication or HbA1C measured by turbidimetry was >= 6%. The participants' education was assessed using an extensive questionnaire and given a score ranging from 1 (no degree at all) to 7 (A-levels + master's degree (or equivalent) or promotion) according to prior research (56) . The effects of education and the significance of different degrees are likely to be culture specific. Fortunately, a recent study examined the effects of education in a population of elderly residents of the city of Leipzig. In this study education operationalised as having a tertiary degree or not was found to be a significant predictor of dementia incidence (57) . This is approximated with a cut-off at a score < 3.6. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted December 16, 2021. ; Participants had to choose their gender in a binary female/male question. Note that the German "Geschlecht" does not differentiate between sex and gender. The lack of a clarification and other options is lamented by the authors. We used the sum-score of the Center for Epidemiological Studies Depression Scale (CES-D) to measure depressive symptoms (58, 59) . For a sensitivity analysis we created a dichotomous variable coded as 1 if participants answered the LSNS questionnaire after March 22 nd , 2020 (1 st SARS-CoV-2 lockdown in Germany). We excluded outliers for our core variables (LSNS-score, adjusted hippocampal volume, cognitive functions). Please see Fig. 1 (Flowchart) for the effect of outlier exclusion on the different models and the supplement for details on handling outliers. To avoid an excessive reduction in sample size due to missing data we performed some limited imputations for missing predictor variables. Please see Sup. 1 for information on our procedures. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted December 16, 2021. ; https://doi.org/10.1101/2021.12.14.21267787 doi: medRxiv preprint All code can be found on https://github.com/LaurenzLammer/socialisolation. The preregistration offers a more extensive description of all our models. To investigate the link between social isolation and our outcomes of interest, we employed LMEs with individual as a random effect. We calculated two models for each hypothesis. In model one we included age and gender as control variables. Model two additionally included education, hypertension, diabetes, depressive symptoms and BMI. In model one the other risk factors are assumed to mediate the effect of social isolation. In model two they are assumed to be confounders (see Sup. 2 for a visualization). To measure the effect of aging, we controlled for baseline age and change in age. Analogously, we differentiated within and between subject effects(60) of social isolation. Furthermore, to test whether participants that are socially more isolated at baseline experienced more pronounced age-related changes, we investigated the interaction effect of baseline LSNS and change in age. Likewise, we . CC-BY-NC 4.0 International license It is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted December 16, 2021. ; https://doi.org/10.1101/2021.12.14.21267787 doi: medRxiv preprint calculated the interaction effect of baseline LSNS and change in LSNS. With this methodology we regressed hippocampal volume, the three cognitive functions, and cortical thickness on baseline LSNS, change in LSNS, and the interaction terms. To measure the overall effect of our predictors of interest, we performed a full-null-model comparison (61) . In addition to standard p-values, we calculated Bayes Factors (BFs) to evaluate the evidence in favour of the full-hypothesis compared to the null-hypothesis (62, 63) . The relative evidence was measured by dividing the BF for the full model by the BF of the null model (64) . Furthermore, we ran mediation analyses to investigate whether the TICS-score mediated the relationship of LSNS and hippocampal volume and to test whether hippocampal volume mediated the association between LSNS and cognitive functions. Specifically, we investigated the indirect path resulting from the regressions of follow-up mediator on baseline LSNS and follow-up dependent variable on baseline mediator. As in the LMEs, we used models one and two with the different sets of control variables. In addition to these pre-registered analyses, we conducted four sensitivity analyses to test the robustness of our results on hippocampal volume and cognitive functions. For the first analysis we added whether participants were tested after the start of lockdown measures to all LMEs. In the second analysis we didn't exclude participants due to the intake of centrally active or cancer medication and cognitive impairment. To probe the reliability of the coefficients for LSNS_change, we ran an analysis excluding all participants with only one timepoint and used standard mean and within score calculation. To test for potential differences in the effect of social isolation between women and men, we divided our dataset by gender and recalculated the frequentist LMEs with both resulting datasets. In order to further investigate the nature of the correlations, we calculated bivariate latent change score (BLCS) models (65) . In these models we simultaneously tested for an effect of baseline social isolation on change in cognitive functions or hippocampal volume and vice versa (see Sup. 3 for a visualization). We sided p-values based on the direction of the predictor/path of interest's regression coefficient and the direction of our pre-defined hypotheses. To side the BFs we sampled 10,000 times from the posterior distribution of our predictor of interest's effect. Then we multiplied the BF by two and the percentage of sampled effects in the direction of our predefined hypotheses. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted December 16, 2021. ; Our threshold for significance for all tests was p < 0.05. To control for multiple hypothesis testing we FDR-corrected families of tests and each individual whole brain analysis (see Sup. 1 for definition of families). BFs of 3 to 10 and BFS of 10 to 30 are commonly considered to be moderate or strong evidence in favour of a hypothesis(66). To evaluate these thresholds in light of multiplicity, we conducted two simulation studies described in Sup.1 that revealed that using a BF threshold of 10.75 rather than 3 would keep α below 5% and that this would not substantially decrease power. To ensure that our continuous predictors are normally distributed, we plotted their histograms. We had to log-transform the CES-D-score to obtain a normal distribution. To rule out major collinearity, we calculated Variance Inflation Factors (VIFs). For LMEs, we dropped the random effect and calculated a linear model to obtain the VIFs. The VIFs did not surpass the threshold of 10(67) in any model. Furthermore, we tested the stability of our LMEs in R by comparing the estimates obtained from the model based on all data with those obtained from models with the levels of the random effects excluded one at a time(68). This revealed the models to be fairly stable. Moreover, we visually controlled the normality of the residuals of these models with both a histogram and a qq-plot. The qq-plots show a heavy-tailed distribution of the residuals in some models. This is only a minor deficit as the models are not intended to make accurate predictions (69) . Fit indices providing further information on the quality of a model fit using structural equation modelling can be found in Sup. 4-5. As suggested by Schermelleh-Engel et al., (70) we report χ 2 and its associated p value, χ 2 /df, RMSEA and its associated confidence interval, SRMR, NNFI, and CFI. Fit index thresholds were surpassed by multiple mediation models. As the BLCS models are saturated, fit indices are uninformative. We performed most analyses using R (R Project for Statistical Computing, V3.6.1, RRID:SCR_001905). For the whole brain analyses we used Matlab (MATLAB, V9.10 (2021a) RRID:SCR_001622). is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted December 16, 2021. ; https://doi.org/10.1101/2021.12.14.21267787 doi: medRxiv preprint We used the package lme4 (R package: lme4, RRID:SCR_015654) to calculate LMEs in R. To obtain reliable p-values, we used the Satterthwaite option from the lmerTest package( R package: lmerTest, RRID:SCR_015656) (71) . In the whole brain analyses we employed the Matlab-toolbox provided by FreeSurfer to calculate vertex-wise LMEs (72) . For mediation analyses and BLCS models we used the sem function from the lavaan package (73) . We calculated BFs for all LMEs in R using the BayesFactor package and the functions posterior and generalTestBF with default priors (74) . VIFs were calculated using the package car (76) . In quantitative studies, despite its importance in shaping the research process and conclusions, e.g. in functional MRI analysis (77) , researchers' influence is often disregarded. In Sup. 1 we offer a brief reflexivity section to make relevant influences on this study transparent and to shortly discuss the value of reflexivity for quantitative science. In total, we included baseline data from 1992 participants and follow-up data from 1409 participants with a mean age of 67 and 68 years, respectively, thereof 46% and 47% women, respectively and a ~6 years mean change in age at follow-up. The sample displayed a prevalence of 60% hypertension, < 20% diabetes and 11-13% had no tertiary education. Descriptive statistics including age range, gender distribution, cardiovascular risk factor prevalence and examination under pandemic conditions of the study sample at baseline and follow-up are shown in Table 1 . is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted December 16, 2021. ; Table 1 Social isolation and hippocampal volume We found that both, stronger baseline social isolation (values for models 1/2: β =< −5. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted December 16, 2021. ; https://doi.org/10.1101/2021.12.14.21267787 doi: medRxiv preprint Overall, stronger baseline social isolation and to a lesser extent increases in social isolation, linked to worse cognitive performance (Table 3, Fig 2) . Specifically, stronger social isolation at baseline significantly predicted lower executive functions (β = −0.026/−0.015 SD/pt, q = 1.0e-07/0.0046) and lower processing speed (β = −0.018/−0.018 SD/pt, q= 1.0e-05/1.2e-04). The link to lower memory (β = −0.014/−0.008 SD/pt, q = 0.002/0. 0775) was strong in model is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted December 16, 2021. ; https://doi.org/10.1101/2021.12.14.21267787 doi: medRxiv preprint Table 2 . CC-BY-NC 4.0 International license It is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted December 16, 2021. ; https://doi.org/10.1101/2021.12.14.21267787 doi: medRxiv preprint Table 3 . CC-BY-NC 4.0 International license It is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted December 16, 2021. ; https://doi.org/10.1101/2021.12.14.21267787 doi: medRxiv preprint Scatterplots with regression lines and 95% confidence intervals for model 1. Asterisks show frequentist levels of significance. The 1 st and 2 nd line show values before and after FDR, respectively. **** p < 0.0001, *** p < 0.001, ** p < 0.01, * p < 0.05. Pie charts show bayesian relative evidences. The green and black arc lengths represent the evidence in favour of the alternative and the null hypothesis, repectively. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted December 16, 2021. ; https://doi.org/10.1101/2021.12.14.21267787 doi: medRxiv preprint LSNS_base, baseline Lubben Social Network Scale; age_base, baseline age; LSNS_change, change in Lubben Social Network Scale; age_change, change in age For the gender variable and for the education variable being female and having at least a tertiary degree were coded as 0, respectively. Betas were standardized by the standard deviations of the dependent and independent variable. In the whole brain analysis, we found a total of four clusters of decreased cortical thickness associated with stronger baseline social isolation in model 1 after FDR correction with an alpha level of 5% (Fig 5) . The clusters were located in the left precuneus and right supramarginal gyrus, superior temporal gyrus and cuneus. Increases in social isolation were linked to decreased cortical thickness in 7 clusters in the right middle and superior frontal is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted December 16, 2021. ; https://doi.org/10.1101/2021.12.14.21267787 doi: medRxiv preprint gyri, orbitofrontal and lateral occipital cortex (Fig 6) . When controlling for further covariates, three of these in the middle/superior frontal and lateral occipital gyrus remained significantly associated with the largest one splitting into two smaller clusters. Table 4 lists these clusters, their locations and sizes. Table 4 . CC-BY-NC 4.0 International license It is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted December 16, 2021. ; https://doi.org/10.1101/2021.12.14.21267787 doi: medRxiv preprint Whole brain analysis of the effect of baseline social isolation on cortical thickness Unstandardized betas are the vertex-wise effect sizes of baseline social isolation in mm/point on the Lubben Social Network Scale corrected for baseline age, change in age, change in social isolation and gender. The first row shows the left hemisphere. The right hemisphere is shown below. First and second column show the lateral and medial view, respectively. The box on the right shows two clusters of reduced cortical thickness associated with social isolation in the left precuneus that were significant after FDR-correction and the F-value of each significant vertex. On the right hemisphere we detected significant clusters after FDR-correction, too. They are located in the supramarginal gyrus and cuneus and are not highlighted in this figure. Whole brain analysis of the effect of change in social isolation on cortical thickness Unstandardized betas are the vertex-wise effect sizes of change in social isolation in mm/point on the Lubben Social Network Scale corrected for baseline age, change in age, baseline social isolation and gender. The first row shows the left hemisphere. The right hemisphere is shown below. First and second column show the lateral and medial view, respectively. The box on the right shows clusters of reduced cortical thickness associated with social isolation in the right superior and middle frontal gyrus, and lateral and medial orbitofrontal cortex that were significant after FDR-correction and the Fvalue of each significant vertex. Additionally, we detected another significant cluster after FDRcorrection in the lateral occipital cortex that is not highlighted in this figure. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted December 16, 2021. ; https://doi.org/10.1101/2021.12.14.21267787 doi: medRxiv preprint Neither the mediation analyses with chronic stress as a mediator for the effect of social isolation on hippocampal volume (n = 62 observations) nor the mediation analyses with hippocampal volume as a mediator for the effect of social isolation on cognitive functions (n = 313-331) yielded any significant results. Due to the requirements of the model design and over 50% missingness in the stress questionnaire the sample sizes of the mediation analyses were gravely diminished. Details on the mediation analyses are provided in Sup. 9. Analyses accounting for a) potential effects of measurements before compared to during the Covid-19 pandemic, b) reducing the exclusion criteria (i.e., not excluding cognitively impaired participants, participants taking centrally active medication and participants with recent cancer treatment) and c) only including participants with two timepoints and using mean and within scores, confirmed the regression coefficients of our models in terms of direction and size (Sups 10-15). Neuroscience has historically neglected sex and gender differences, predominantly resulting in increased misdiagnoses of and relatively worse treatments for women (78) . Therefore, we recalculated analyses in gender-stratified samples (n female = 1125 observations, n male = 1105 observations) to test for differences in the effects of social isolation (Sup 16). No clear pattern of difference emerged between women and men. A minor observable difference was that the interaction of baseline social isolation with change in age on hippocampal volume was more pronounced in men. The bivariate latent change score models did not result in solid evidence for directionality. With a p-value of 0.019 and an effect size of −0.015 mm 3 there is very weak evidence that lower baseline social isolation predicted a decrease in executive functions (Sup. 17). As in the mediation analyses, the design of the BLCS resulted in smaller sample sizes (n = 333-548). In this pre-registered study, we investigated the associations of social isolation with brain structure and cognition in a large cognitively healthy mid-to late-life longitudinal sample. In line with our pre-specified hypotheses, we showed a significant link between stronger baseline social isolation and increases in social isolation over the course of ~ 6 years and smaller hippocampal volumes. Both predictors had an effect size per point on the LSNS comparable to a two and a half-month difference in baseline age in this age range. Simply put, assuming that if everything else remained stable, the difference between having one or 3-4 . CC-BY-NC 4.0 International license It is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted December 16, 2021. ; https://doi.org/10.1101/2021.12.14.21267787 doi: medRxiv preprint close and supportive friends is comparable to a one-year difference in hippocampal aging. Furthermore, we found significant associations of stronger baseline social isolation with lower executive functions, memory and processing speed. The link to executive functions was particularly strong with an effect size larger than a one-year difference in baseline age. For increases in social isolation, confidence intervals were wider but effect sizes, except for executive functions, were similar in magnitude to that of baseline social isolation. Moreover, there was an interaction effect of baseline social isolation with change in age on hippocampal volumes indicating accelerated brain aging in more isolated individuals. In multiple sensitivity analyses we showed the robustness of these findings. Neither applying less exclusion criteria, only including participants with two timepoints nor controlling for the impact of the ongoing pandemic changed our results substantially. Moreover, we found clusters of decreased cortical thickness in the cuneus, precuneus, lateral occipital cortex, supramarginal gyrus, orbitofrontal cortex and superior and middle frontal gyrus associated with social isolation cross-sectionally or longitudinally. Mediation analysis revealed no significant effects. Our findings indicate that social isolation contributes to grey matter loss in the hippocampus, a focal point of atrophy in mild cognitive impairment (79) and dementia (80) . Notably, not only baseline social isolation (a between-subject effect) but also change in social isolation (a within-subject effect) significantly predicted hippocampal volume. Through the employment of statistical LMEs, we were able to distinguish and study effects at these different levels (60, 81) and the design helped us to avoid fallacious inferences from single level data (82) to which simple linear regressions would have been susceptible. Specifically for the study of social isolation as a risk factor for dementia, it is crucial to disentangle between-and within-subject effects. Social isolation has both been described as a trait (83), implying it to be an invariant between-subject characteristic and as a potential target for interventions (84) , implying it to be a modifiable within-subject effect. The finding of a significant within-subject effect of change in social isolation therefore offers hope for modifiability as it implies that the observed associations are not (exclusively) the effect of an invariant trait. Thus, our data point towards that reducing social isolation could help to maintain hippocampus integrity in aging. However, this assumes a causal effect of social isolation. As associations with social isolation could also have resulted from reverse causation through health selection, i.e. that participants . CC-BY-NC 4.0 International license It is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted December 16, 2021. ; with accelerated brain aging are more likely to become socially isolated, this assumption needs careful consideration. On the one hand, our interaction models designed to test the temporality of the effect, provided evidence for an interaction of baseline social isolation and change in age on hippocampal volume, pointing towards a detrimental effect of social isolation. Bayesian statistics, on the other hand, imply the absence of an interaction effect for all other dependent variables and the bivariate latent change score models barely provide evidence in favour of causality in the hypothesized direction. This inconclusiveness might result from our reduced follow-up sample size, the still relatively narrow change in time compared to the adult lifespan and limited number of time points in the current study. For example, data from the English Longitudinal Study of Aging from > 6000 older adults measured at up to 6 two-year intervals supports the assumed causality of social isolation with regards to memory performance (85) . Moreover, the presence of considerable effect sizes and the high statistical confidence in these estimates on multiple outcomes in this healthy sample without cognitive impairment speaks against the competing hypothesis of reverse causality through health selection and in favour of a causal role of social isolation. Furthermore, the lack of any strong increase in effect size when including health-impaired participants corroborates this interpretation. Baseline social isolation, and to a lesser extent, change in isolation, were significantly associated with cognitive performance, i.e. executive functions, processing speed and memory, all of which undergo decline in (pathological) aging (86) . Again, our results thus imply a detrimental role of social isolation on cognitive functions. We could however not observe that social isolation lowered memory performance through reductions in hippocampal volume, a hypothesis raised by considerations of the central role of the hippocampus in memory (87, 88) . Similarly, we could not find evidence that social isolation affected hippocampal volume through higher chronic stress measured with questionnaires, a hypothesis put forward by the stress buffering theory (36) . However, these latter analyses suffered from small sample sizes and a limited number of timepoints. Overall, comparing our brain morphometric results with those of existing cross-sectional studies on social isolation, detected brain regions coincide. A rather small-sampled study did not find a link between social isolation and grey matter volumes (89) is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted December 16, 2021. ; https://doi.org/10.1101/2021.12.14.21267787 doi: medRxiv preprint lobe) (27) and Blumen and Verghese (hippocampus, precuneus, superior frontal gyrus, medial frontal gyrus) (29) found decreased volumes in regions we detected, too. Several of the cortical regions identified in our study (precuneus, orbitofrontal cortex) belong to the pattern of exacerbated regional atrophy found in Alzheimer's disease. Furthermore, we detected regions known for increased cortical thinning in the healthy process of aging (cuneus, lateral occipital cortex, inferior frontal gyrus) and both in healthy and pathological aging (supramarginal gyrus, medial frontal gyrus) (90, 91) . This indicates an aggravating role of social isolation in cortical thinning that may contribute to normal and accelerated brain aging processes. However, the findings of reduced cortical thickness must be interpreted cautiously due to the limited consistency between cross-sectional and longitudinal effects and the exploratory approach of whole brain analyses. Contrasting our frequentist and Bayesian measures of significance, a noticeable discrepancy between small BFs (indicating no evidence in favour of the alternative hypothesis) and pvalues < 0.05 (indicating evidence to reject the null hypothesis) is observable for some models of change in social isolation. While BFs are generally more conservative than pvalues (92) , this mismatch might additionally result from the software's standardization procedure that standardizes raw effect sizes by both the dependent and independent variable's SDs (74) . While the effect size per point on the LSNS are mostly very similar, the smaller SD of change in LSNS score makes the standardized effect sizes much smaller than those of baseline LSNS. Thus, the lower BFs for change in social isolation are in part due to these limitations, which urges us to interpret them more carefully. This can also be seen in the results of our power simulation shown in Sups. [18] [19] . A limitation of this study is its uncertain generalizability to the general population because the sample was probably affected by selection and attrition bias common to longitudinal studies (93) . Along these lines, our population study represents a WEIRD sample (i.e., western, educated, industrialised, rich, democratic)(94) which might skew our understanding of how social isolation affects brain health (95) . Additionally, attrition might have caused some degree of conflation of different effect levels so that within-subject effects of participants with only one timepoint are treated as between-subject effects. A further limitation are ceiling effects in the memory and processing speed scores in healthy adults, is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted December 16, 2021. ; https://doi.org/10.1101/2021.12.14.21267787 doi: medRxiv preprint potentially limiting the sensitivity to detect subtle differences. Moreover, covariance of social isolation with other variables such as hypertension or diabetes could have influenced the results. However, note that all VIFs were acceptable. Lastly, we must stress that we investigated cognitive functions and brain morphometry and not diagnosed dementia. Henceforth, all inferences from our results on dementia must be made with caution. This pre-registered large-scale population neuroimaging analysis adds robust support to the view that social isolation is associated with accelerated brain aging and cognitive decline in non-demented adults in mid-to late-life. Our findings further imply that social contact protects from detrimental processes and thereby preserves brain structure and function. Henceforth, targeting social isolation through tailored strategies might contribute to maintaining brain health into old age. While we see evidence converging on social isolation as a causal risk factor for dementia and cognitive decline, future neuroimaging studies should pay particular attention to questions of temporality in their design to clear up remaining uncertainties. Intervention studies will be the gold standard to provide evidence with regards to the causal role and effect size of social isolation. Yet, multidomain interventions for dementia prevention justifiably become the norm (96) (97) (98) , so that effects of reduced social isolation must be investigated as a likely contribution to an aggregate effect. Illuminating the mechanistic underpinnings of the association should be another focus for future research. Promising approaches to answer this research question could be interventions specifically targeting one of the hypothesized detrimental processes in isolated individuals and mediation analyses of multi-wave studies with larger sample sizes. Studies should also prioritise obtaining reliable proxies for the hypothesized mediators. As elevated cortisol levels, in line with the stress-buffering hypothesis, may exert detrimental effects on cognition and contribute to AD pathology(99), using hair cortisol, a reliable measure of chronic stress(100,101), could be a promising choice to further investigate this proposed mechanism. In light of the relevance of social isolation for cognitive and general health and wellbeing (102) , its pervasiveness in the elderly population of the global north (7,103,104) is alarming. Physical distancing measures have caused an unprecedented rise in the attention to the impact of social isolation(105) but social isolation has been a grave problem before Covid-19 and it will remain a central public health concern thereafter. Existing and future research on the role of social isolation in health and disease should provide guidance for the . CC-BY-NC 4.0 International license It is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted December 16, 2021. ; https://doi.org/10.1101/2021.12.14.21267787 doi: medRxiv preprint The Global Impact of Dementia. An Analysis of Prevalence, Incidence, Cost and Trends Global burden of 369 diseases and injuries in 204 countries and territories, 1990-2019: a systematic analysis for the Global Burden of Disease Study Critical Appraisal of Amyloid Lowering Agents in AD Failure to demonstrate efficacy of aducanumab: An analysis of the EMERGE and ENGAGE trials as reported by Biogen A Systematic Review and Meta-Analysis of the Effectiveness of Acetylcholinesterase Inhibitors and Memantine in Treating the Cognitive Symptoms of Dementia Donepezil for dementia due to Alzheimer's disease. Cochrane database Syst Rev Dementia prevention, intervention, and care: 2020 report of the Lancet Commission Social relationships and risk of dementia: A systematic review and metaanalysis of longitudinal cohort studies The Association between Social Engagement, Loneliness, and Risk of Dementia: A Systematic Review and Meta-Analysis Social isolation, cognitive reserve, and cognition in healthy older people Social relationships and cognitive decline: a systematic review and meta-analysis of longitudinal cohort studies The impact ofsocial activities, social networks, social support and social relationships on the cognitive functioning of healthy older adults: A systematic review. Syst Rev Tracking pathophysiological processes in Alzheimer's disease: an updated hypothetical model of dynamic biomarkers Spatial patterns of neuroimaging biomarker change in individuals from families with autosomal dominant Alzheimer's disease: a longitudinal study Neuroimaging markers for the prediction and early diagnosis of Alzheimer's disease dementia Gray matter age prediction as a biomarker for risk of dementia MRI cortical thickness biomarker predicts AD-like CSF and cognitive decline in normal adults Secondary prevention of Alzheimer's dementia: Neuroimaging contributions. Alzheimer's Res Ther Exercise in Adults with Mild Memory Problems (EXERT) Brain volumes and cortical thickness on MRI in the Finnish Geriatric Intervention Study to Prevent Cognitive Impairment and Disability (FINGER) White matter structure in loneliness: Preliminary findings from diffusion tensor imaging Social network diversity and white matter microstructural integrity in humans The default network of the human brain is associated with perceived social isolation Association of social engagement with brain volumes assessed by structural MRI Gray mattervolume covariance networks, social support, and cognition in older adults Gray matter volume covariance networks associated with social networks in older adults Structural Brain Correlates of Loneliness among Older Adults Variability in Brain Structure and Function Reflects Lack of Peer Support Population variability in social brain morphology for social support, household size and friendship satisfaction Threeyear changes in leisure activities are associated with concurrent changes in white matter microstructure and perceptual speed in individuals aged 80 years and older Changes in brain volume and cognition in a randomized trial of exercise and social interaction in a community-based sample of non-demented chinese elders What Has Replication Ever Done for Us? Insights from Neuroimaging of Speech Perception Social ties and mental health Use it or lose it: Engaged lifestyle as a buffer of cognitive decline in aging? Examining the role of memory in social cognition Social network types and the health of older adults: Exploring reciprocal associations Rethinking the health selection explanation for health inequalities Do High Mental Demands at Work Protect Cognitive Health in Old Age via Hippocampal Volume? Results From a Community Sample The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies Best practices in data analysis and sharing in neuroimaging using MRI The LIFE-Adult-Study: objectives and design of a population-based cohort study with 10,000 deeply phenotyped adults in Germany Critical ages in the life course of the adult brain: Nonlinear subcortical aging Within-subject template estimation for unbiased longitudinal image analysis Reliability and statistical power analysis of cortical and subcortical FreeSurfer metrics in a large sample of healthy elderly Qoala-T: A supervised-learning tool for quality control of FreeSurfer segmented MRI data Performance of an abbreviated version of the Lubben Social Network Scale among three European community-dwelling older adult populations Loneliness, social isolation and social relationships: What are we measuring? A novel framework for classifying and comparing tools The effects of intracranial volume adjustment approaches on multiple regional MRI volumes in healthy aging and Alzheimer's disease Higher body mass index in older adults is associated with lower gray matter volume: Implications for memory performance Higher body mass index is associated with reduced posterior default mode connectivity in older adults. Hum Brain Mapp Trierer Inventar zum chronischen Stress : TICS. Göttingen : Hogrefe Messung des sozioökonomischen Status in der Studie Education as protector against dementia, but what exactly do we mean by education? A Self-Report Depression Scale for Research in the General Population Allgemeine Depressionsskala (ADS). 2. Göttingen: Hogrefe A simple method for distinguishing within-versus betweensubject effects using mixed models Generalized linear mixed models: a practical guide for ecology and evolution The importance of proving the null Using Bayes factor hypothesis testing in neuroscience to establish evidence of absence Model comparison in ANOVA Developmental cognitive neuroscience using latent change score models: A tutorial and applications Classical and modern regression with applications Data Analysis Using Regression and Multilevel/Hierarchical Models Evaluating the Fit of Structural Equation Models: Tests of Significance and Descriptive Goodness-of-Fit Measures LmerTest: Tests in linear mixed effects models. R Packag Version Statistical analysis of longitudinal neuroimage data with Linear Mixed Effects models {lavaan}: An R Package for Structural Equation Modeling Default Bayes Factors for Model Selection in Regression Spatiotemporal linear mixed effects modeling for the mass-univariate analysis of longitudinal neuroimage data An {R} Companion to Applied Regression [Internet]. Third. Thousand Oaks {CA}: Sage Variability in the analysis of a single neuroimaging dataset by many teams Considering sex as a biological variable will require a global shift in science culture Hippocampal and entorhinal atrophy in mild cognitive impairment: prediction of Alzheimer disease Presymptomatic hippocampal atrophy in Alzheimer's disease. A longitudinal MRI study Ecological Correlations and the Behavior of Individuals Home alone: A population neuroscience investigation of brain morphology substrates Primary prevention of dementia: from modifiable risk factors to a public brain health agenda? Social Isolation and Memory Decline in Laterlife Cognitive Aging: Progress in Understanding and Opportunities for Action Memory, navigation and theta rhythm in the hippocampalentorhinal system Memory and the hippocampus: a synthesis from findings with rats, monkeys, and humans No strong evidence that social network index is associated with gray matter volume from a data-driven investigation The effects of aging and Alzheimer's disease on cerebral cortical anatomy: Specificity and differential relationships with cognition Brain atrophy in Alzheimer's Disease and aging Calibration of p Values for Testing Precise Null Hypotheses A systematic literature review of attrition between waves in longitudinal studies in the elderly shows a consistent pattern of dropout between differing studies The weirdest people in the world? open datasets for human connectomics research: Considerations for reproducible and responsible data use World-Wide FINGERS Network: A global approach to risk reduction and prevention of dementia de -Study protocol of a pragmatic multi-center cluster-randomized controlled prevention trial against cognitive decline in older primary care patients High Cortisol and the Risk of Dementia and Alzheimer's Disease: A Review of the Literature Technical and clinical aspects of cortisol as a biochemical marker of chronic stress Hair cortisol, stress exposure, and mental health in humans: a systematic review Social Isolation and Loneliness in Older Adults: Opportunities for the Health Care System Social factors and the prevalence of social isolation in a population-based adult cohort Change in the prevalence of social isolation among the older population from 2010 to 2016: A repeated crosssectional comparative study of Japan and England urgently needed development and evaluation of tailored strategies against social isolation and its detrimental effects. These should address social isolation both through intervention strategies on the individual but also societal level, leveraging values like solidarity and communality.