A Meta-analysis of Loneliness and Risk of Dementia using Longitudinal Data from >600,000 Individuals

Martina Luchetti; Damaris Aschwanden; Amanda A Sesker; Xianghe Zhu; Páraic S O’Súilleabháin; Yannick Stephan; Antonio Terracciano; Angelina R Sutin

doi:10.1038/s44220-024-00328-9

. Author manuscript; available in PMC: 2025 Nov 1.

Published in final edited form as: Nat Ment Health. 2024 Oct 9;2(11):1350–1361. doi: 10.1038/s44220-024-00328-9

A Meta-analysis of Loneliness and Risk of Dementia using Longitudinal Data from >600,000 Individuals

Martina Luchetti ¹, Damaris Aschwanden ^2,³, Amanda A Sesker ⁴, Xianghe Zhu ^5,⁶, Páraic S O’Súilleabháin ^7,⁸, Yannick Stephan ⁹, Antonio Terracciano ¹⁰, Angelina R Sutin ¹¹

PMCID: PMC11722644 NIHMSID: NIHMS2029056 PMID: 39802418

Abstract

Loneliness is one critical risk factor for cognitive health. Combining data from ongoing aging studies and the published literature, we provided the largest meta-analysis on the association between loneliness and dementia (k = 21 samples, N = 608,561) and cognitive impairment (k = 16, N = 103,387). Loneliness increased risk for all-cause dementia (HR = 1.306, 95% CI [1.197,1.426]), Alzheimer’s disease (HR = 1.393, 95% CI [1.290,1.504]; k = 5), vascular dementia (HR = 1.735, 95% CI [1.483,2.029]; k = 3), and cognitive impairment (HR = 1.150, 95% CI [1.113,1.189]). The associations persisted when models controlled for depression, social isolation, and/or other modifiable risk factors for dementia. The large heterogeneity across studies was partly due to differences in loneliness measures and ascertainment of cognitive status. Results underscored the importance to further examine type/sources of loneliness and cognitive symptoms to develop effective interventions that reduce the risk of dementia.

Introduction

Loneliness is a distressing feeling that impacts the health and well-being of older adults^1,2. It differs from being alone or isolated; people can feel lonely even when with others if the quality or quantity of their social connections does not meet what is needed or desired^3,4. Many observational studies find that individuals who feel lonely (versus not lonely) have a higher risk for developing dementia and other cognitive impairments^5–10, even when accounting for measures of social isolation (e.g., few social contacts), clinical (e.g., cardiovascular disease), behavioral (e.g., physical inactivity), and genetic risk factors for dementia^6,9. Null or mixed results have also been reported^11–13, with a significant association only for subgroups of participants (e.g., male individuals)^12,14. In the past years, and particularly after the social restrictions due to the COVID-19 pandemic, there has been an increasing interest in the consequences of loneliness, with numerous new publications and attempts to identify risks for cognitive health^15–18. In a pre-pandemic meta-analysis, Lara et al.¹⁵ pooled results from 8 independent studies (n~30,000) and estimated a Relative Risk of 1.26 (95%CI = 1.14,1.40) to develop all-cause dementia or Alzheimer’s disease for individuals who feel lonely. The moderate heterogeneity (I² = 23%) across studies indicated potential differences due to design (e.g., follow-up length), measure of loneliness, and control variables. Since their synthesis, others have meta-analyzed the published literature^16,18,19. These latter studies, however, applied somewhat different criteria for study selection^18,19 and assessed loneliness risk together with other functional and structural aspects of social health (e.g., social isolation, support, marital status)^16,17. This approach confounds the association of loneliness, which is independent of the availability of social relationships^6,8. Further, recent publications^9,13,20,21 were not included in previous meta-analyses.

The present meta-analysis builds on and expands this work in three fundamental ways. First, in addition to the association between loneliness and risk for all-cause dementia, we evaluate risk for two cause-specific dementias: Alzheimer’s disease (AD) and vascular dementia (VaD). Second, we examine the association between loneliness and cognitive impairment, not dementia (CIND), or non-specific impairments in one or more cognitive functions (CI), that may precede the onset of the disease²². The one meta-analysis¹⁹ that examined the association between loneliness and AD and CIND/CI included only three studies for each of these outcomes. Third, to gather as much data as possible, we combined results from the published literature with a coordinated analysis of ongoing aging cohort studies that assessed loneliness and cognition over time. This allowed us to examine sources of heterogeneity across studies, including loneliness measures, dementia ascertainment methods, and sample characteristics. Previous meta-analyses^15,18 included 10 or fewer studies, which was too few to identify possible moderators with meta-regression. Here, we aimed to provide a comprehensive, up-to-date analysis of loneliness-associated risk for dementia-related outcomes in the general population.

Results

This meta-analysis addressed two questions: Q1: Is loneliness related to an increased risk of incident all-cause dementia and cause-specific dementias, such as AD and VaD? and Q2: Is loneliness related to an increased risk of incident CIND/CI? The questions were developed utilizing the PICOS framework: P (population) = cognitively unimpaired middle-aged and older adults at baseline; I (intervention) = no intervention, assessment of loneliness; C (comparison) = no control group/comparison; O (outcome) = risk of incident dementia (Q1, primary outcome), including all-cause dementia, AD and VaD; risk of incident CIND/CI (Q2, secondary outcome); and S (study design) = longitudinal observational studies. The protocol was pre-registered in PROSPERO (CRD42022331818). Deviations from the pre-registration are reported in Supplementary Table S1²³.

Coordinated Analysis of Aging Cohort Studies

A search of the Gateway to Global Aging (https://g2aging.org) identified seven cohort studies that included loneliness and repeated assessments of cognitive status: the Health and Retirement Study (HRS), the English Longitudinal Study of Ageing (ELSA), the Survey of Health, Ageing and Retirement in Europe (SHARE), the Irish Longitudinal Study on Ageing (TILDA), the Mexican Health and Aging Study (MHAS), the Korean Longitudinal Study of Aging (KLoSA), and the China Health and Retirement Longitudinal Study (CHARLS). In addition, we were aware from our previous work that the Household, Income and Labour Dynamics in Australia (HILDA) survey had relevant data to be included in the analysis (see Supplementary Note S1 for details about each study).

A coordinated analysis that used the same analytical approach (Cox regression) within each cohort was used to estimate the relation between loneliness and risk (hazard ratio) of dementia or CIND/CI using harmonized variables (see Supplementary Tables S2–S4). Cognitive status was classified according to previously validated methods in each cohort (see Methods and Supplementary Table S3). Table 1 summarizes the design, measures, and results for each cohort. The association between loneliness and risk for dementia was tested in HRS, ELSA, SHARE, MHAS, and KLoSA. Risk for CIND was tested in HRS, ELSA, MHAS, and KLoSA, and risk for CI (i.e., performance equal or below 1.5 SD of the age-graded mean of the sample on validated cognitive tasks) was further assessed in SHARE (2 samples), CHARLS and HILDA. For TILDA, we performed analysis for CIND, but not dementia, as there were only 0.4% (n = 19) incident cases of dementia over the follow-up. Loneliness was associated with an increased risk of incident dementia and CIND/CI in all cohorts, except CIND in MHAS. As shown in Extended Data Table 1, risk estimates were reduced in size when controlling for depressive symptoms and social isolation, which are known to be closely related to loneliness^24,25. The associations were mostly unchanged when controlling for modifiable risk factors (diabetes, hypertension, obesity) for dementia²⁶ and when excluding participants younger than 50, who were less likely to develop dementia or CIND/CI. The control variables were selected based on the availability of data across cohorts and the published literature.

Table 1.

Individual Cohort Studies: Summary of study design, measures, and results

Study / Country	Loneliness Measure	Cognitive Measure / Cognitive Status Classification	Neuro-functional Assess.	Outcome of Interest	Total N	Incident Cases, N (%)	Loneliness, M (SD)	Age, M (SD), range	Female individuals, %	Education, M (SD)	Follow-up, M (SD), range	Model 1 HR (95% CI)	Model 2 HR (95% CI)
HRS / US	3-item UCLA (continuous; score 1–3)	TICSm / validated cut-off	No	Dem	12,217	1,335 (10.9)	1.47 (0.53)	67.33 (9.92), 40–99	60.2	12.88 (2.88)	9.94 (4.57), 2–16	1.175^*** (1.116, 1.237)	1.110^*** (1.046, 1.178)
HRS / US	3-item UCLA (continuous; score 1–3)	TICSm / validated cut-off	No	CIND	10,079	3,603 (35.7)	1.44 (0.53)	66.46 (9.63), 40–98	60.6	13.29 (2.59)	9.09 (4.83), 2–16	1.076^*** (1.041, 1.112)	1.011 (1.974, 1.049)
ELSA / UK	3-item UCLA (continuous; score 1–3)	IQCODE and self-reported doctor diagnosis; TICSm / validated cut-off	No	Dem	6,771	442 (6.5)	1.36 (0.49)	64.84 (9.66), 40–90+	56.9	2.51 (2.24)	11.35 (5.55), 1–19	1.190^*** (1.090, 1.298)	1.203^** (1.077, 1.343)
ELSA / UK	3-item UCLA (continuous; score 1–3)		No	CIND	6,349	937 (14.8)	1.35 (0.49)	64.31 (9.46), 40–90+	56.7	2.55 (2.24)	11.12 (5.45), 1–19	1.179^*** (1.107, 1.256)	1.140^** (1.056, 1.230)
SHARE Dementia Sample / Europe (15 countries)	3-item UCLA (continuous; score 1–3)	self/proxy- reported doctor diagnosis / yes/no	No	Dem	35,531	1,887 (5.3)	1.24 (0.41)	64.82 (9.70), 40–101	56.9	2.76 (1.46)	7.81 (3.32), 1–11	1.201^*** (1.115, 1.249)	1.102^*** (1.045, 1.162)
SHARE Sample 1 / Europe (12 countries)	1-item (ordinal; score 1–4)	memory and verbal fluency / 1.5 SD equal or below age-graded mean	No	CI	15,028	969 (6.4)	1.47 (0.74)	62.94 (9.71), 40–98	55.9	2.60 (1.52)	10.74 (5.78), 1–18	1.139^*** (1.078, 1.203)	1.092^* (1.013, 1.178)
SHARE Sample 2 / Europe (15 countries)	3-item UCLA (continuous; score 1–3)	memory and verbal fluency / 1.5 SD equal or below age-graded mean	No	CI	27,673	1,078 (3.9)	1.23 (0.41)	63.99 (9.67), 40–99	57.0	2.79 (1.44)	7.53 (3.57), 1–11	1.181^*** (1.124, 1.242)	1.063 (0.989, 1.143)
TILDA / Ireland	3-item UCLA (continuous; score 1–3)	MMSE / validated cut-off	No	CIND	4,833	149 (3.1)	1.36 (0.49)	62.53 (8.58), 50–80+	54.9	3.88 (1.57)	6.90 (1.97), 2–8	1.315^*** (1.147, 1.507)	1.174^* (1.000, 1.379)
MHAS / Mexico	1-item (categorical/binary)	*see methods for list of measures / data-driven classification by the study team	Yes	Dem	11,672	1.255 (10.8)	31.8%	59.87 (9.61), 40–105	58.1	4.75 (4.40)	10.76 (5.28), 2–15	1.213^*** (1.081, 1.361)	0.931 (0.780, 1.111)
MHAS / Mexico	1-item (categorical/binary)		Yes	CIND	7,349	2,651 (36.1)	29.3%	58.85 (9.41), 40–98	57.2	5.29 (4.58)	8.60 (5.90), 2–15	1.063 (0.977, 1.156)	0.992 (0.877, 1.121)
KLOSA / Korea	1-item (ordinal; score 1–4)	MMSE / validated cut-off	No	Dem	7,991	1,526 (19.1)	1.33 (0.64)	60.25 (10.11), 45–93	54.4	2.07 (1.07)	10.70 (4.43), 2–14	1.163^*** (1.115, 1.212)	1.044 (0.983, 1.109)
KLOSA / Korea	1-item (ordinal; score 1–4)	MMSE / validated cut-off	No	CIND	6,266	2,485 (39.7)	1.28 (0.59)	58.20 (9.33), 45–88	51.0	2.26 (1.06)	9.45 (4.80), 2–14	1.133^*** (1.095, 1.172)	1.094^* (1.046, 1.145)
CHARLS / China	1-item (ordinal; score 1–4)	*see methods for list of measures / 1.5 SD equal or below age-graded mean	No	CI	8,742	1,217 (13.9)	1.43 (0.86)	57.26 (8.76), 40–101	47.2	1.63 (0.81)	6.99 (2.44), 2–9	1.105^*** (1.051, 1.162)	1.023 (0.949, 1.103)
HILDA / Australia	1-item (ordinal; score 1–7)	symbol digit t and backward digit span / 1.5 SD equal or below age-graded mean	No	CI	2,963	26 (1.0)	2.77 (1.95)	53.60 (10.21), 40–89	54.3	2.64 (1.81)	14.17 (1.53), 11–15	1.421^* (1.014, 1.992)	1.013 (0.633, 1.623)

Open in a new tab

Study abbreviations: HRS = Health and Retirement Study; ELSA = English Longitudinal Study of Ageing; SHARE = Study of Health, Ageing and Retirement in Europe; TILDA = The Irish Longitudinal Study on Ageing; MHAS = Mexican Health and Aging Study; KLoSA = Korean Longitudinal Study of Aging; CHARLS = China Health and Retirement Longitudinal Study; and HILDA = Household, Income and Labour Dynamics in Australia. N = number of participants/incident cases; M = Mean; SD = Standard Deviation; HR = Hazard Ratio; 95% CI = 95% Confidence Interval (lower and upper limits). Cognitive Measures: TICSm = Modified Telephone Interview for Cognitive Status; IQCODE = Informant Questionnaire on Cognitive Decline in the Elderly; MMSE = Mini-Mental State Examination. Outcomes: Dem = dementia (all-cause); CIND = cognitive impairment, no dementia; CI = non-specific cognitive impairment (i.e., performance equal or below 1.5 SD of the age-graded mean of the sample on validated cognitive tasks). Cox regression models were performed within each cohort. Two-tailed p-values < .05 were used for all analyses with no adjustment for multiple testing. Model 1 controlled for age, sex, education, race and/or ethnicity; information on race/ethnicity was available only in HRS (15.9% African Americans or other races; 7.5% Hispanic/Latino), ELSA (1.3% non-whites), and HILDA (1.5% Aboriginal or Torres Strait Islander origin). Model 2 additionally controlled for depressive symptoms, social isolation, and modifiable risk factors for dementia; details on measures and harmonization across samples are provided in Supplementary Information (Table S2–S4). Please note that sample sizes were reduced for Model 2 due to missing data on control variables (on average, a 20% reduction across samples). Additional analyses controlling for depression and isolation separately from modifiable risk factors are reported in Extended Data Table 1. Education was reported in years in HRS and MHAS; on a scale from 0 (no qualification) to 6 (degree) in ELSA; from 0 (pre-primary level of education) to 6 (second stage of tertiary education) in SHARE; from 1 (none/some primary) to 7 (postgraduate/higher degree) in TILDA; from 1 (<11 years) to 7 (postgraduate; masters or doctorate) in HILDA; from 1 (elementary school or lower) to 4 (college university or higher) in KLoSA and CHARLS. In each sample (except MHAS), loneliness scores were standardized before the analysis so that one unit corresponded to a 1 SD difference. Please note that for studies that included a classification for both dementia and CIND, the analysis for CIND excluded participants who had CIND at baseline and those who developed dementia before CIND over the follow-up.

^***

p ≤.001

^**

p <.01

p <.05.

Meta-analysis

We followed the Meta-analyses Of Observational Studies in Epidemiology (MOOSE) recommendations²⁷ (see Supplementary Table S5 for a checklist). Fig. 1 summarizes the screening procedure; a breakdown of the database searches is in Extended Data Table 2. The initial literature search revealed 1,035 records: 662 abstracts via PubMed, 80 via Web of Science, 122 via APA PsycNet, and 171 via Science Direct. Four additional articles were identified through screening the reference list of previous reviews. The search was updated during the review process and an additional 234 records were identified across the databases. After title, keyword, and abstract screening, 49 full-text articles were extracted. Of these, 22 were eligible for the meta-analysis; 5 articles were selected for sensitivity analyses. No records were identified through pre-print archives. Details on selection and reasons for exclusion are in Supplementary Table S6.

Fig. 1: — Outcomes of interest: all-cause dementia; AD = Alzheimer’s disease; VaD = vascular dementia; CIND = cognitive impairment, no dementia; CI = non-specific cognitive impairment. Study abbreviations: HRS = Health and Retirement Study; ELSA = English Longitudinal Study of Ageing; SHARE = Study of Health, Ageing and Retirement in Europe.

Extended Data Table 3 and 4 provide a summary of the selected publications. We considered any study that assessed dementia based on clinical data, medical or death records, cognitive/neuropsychological evaluations, self or proxy reports of a doctor diagnosis. The diagnostic process, however, likely did not consider recent in-vivo biomarkers that can better diagnose cause-specific dementia, particularly Alzheimer’s disease²⁸. When assessing risk of CIND/CI, we included all studies independently of the causes of cognitive impairment. The analysis encompassed studies that assessed CIND/CI or Mild Cognitive Impairment (MCI), a more stringent classification originally identified as a precursor of AD²².

Primary outcome = Dementia

For all-cause dementia, 21 samples were included in the meta-analysis (total N = 608,561). These samples included 16 samples identified from 13 articles through the search of the literature^{7,9–11,13,14,20,29–34} and 5 of the individual cohort studies. Thirteen out of 21 studies reported a significant association between loneliness and dementia risk (Fig. 2); the pooled estimate indicated that feeling lonely increased risk for dementia by 31% (HR = 1.306 95% CI [1.197,1.426], p <.001). There was large heterogeneity across studies (Q = 117.91, df = 20, p <.001; I² = 87.04%; tau² = .02). The leave-one-out analysis (Extended Data Table 5) indicated that the meta-analytic effect was not dependent on any single study; heterogeneity was reduced when excluding one study by Sutin et al.⁹, which was the largest study included in the analysis (n > 450,000, while most studies included < 10,000 participants).

Fig. 2: — The plot displays individual study estimates (21 samples, 608,561 participants) and the average random-effects (RE) model for all-cause dementia. Effect sizes are hazard ratios (HR) with 95% confidence intervals (95% CI). Upper limits of the 95% CIs exceeding 4.0 are not shown. Study abbreviations: H70 Study = Gothenberg H70 Birth Cohort Study; LEILA = Leipzig Longitudinal Study of the Aged; LRGSTUA = Neuroprotective Model for Healthy Longevity among Malaysian Older Adults Towards Using Ageing; RS = Rotterdam Study; SNAC-K = Swedish National Study on Aging and Care in Kungsholmen. Selected cohorts: HRS = Health and Retirement Study; ELSA = English Longitudinal Study of Ageing; SHARE = Study of Health, Ageing and Retirement in Europe; MHAS = Mexican Health and Aging Study; and KLoSA = Korean Longitudinal Study of Aging.

Sensitivity Analysis.

In a follow-up analysis, we included an additional study published by Rolandi et al.³⁵. This study used a competitive risk factor model to estimate risk of dementia, which made it less comparable to other studies in the meta-analysis. The association between loneliness and dementia was similar when including this study (HR = 1.297, 95% CI [1.189,1.415]). We further conducted two (unregistered) analyses: The first analysis included results from fully adjusted models, which controlled for depressive symptoms, social isolation, and/or modifiable risk factors for dementia (k = 18, N = 559,890; three studies did not control for these factors; see Data Extraction File). This analysis indicated that loneliness was still associated with risk of dementia: the effect size was attenuated but not due entirely to these factors (HR = 1.189, 95% CI [1.101,1.285]; Extended Data Fig. 1, panel a). The second analysis restricted results to samples that only included adults 50 or older (N = 489,777); the effect size was virtually the same as in the main analysis (HR = 1.305, 95% CI [1.194,1.428]).

Publication Bias.

The Egger’s test was 1.04 (p = 0.310), indicating no significant evidence of publication bias (Extended Data Fig. 2, panel a). The bias-corrected trim and fill estimate was similar to that obtained in the main analysis (HR = 1.297, 95% CI [1.189,1.414]).

Meta-regressions.

Three significant effects were found (Table 2): (1) Studies that examined presence/absence of loneliness found a stronger association (k = 16; HR = 1.406, 95% CI [1.209,1.635]) compared to studies that used an ordinal/continuous scale (k = 5; HR = 1.182, 95% CI [1.154,1.211]). (2) The association was stronger in studies that used a clinical diagnosis (k = 13; HR = 1.475, 95% CI [1.256,1.732]) compared to studies that used performance-based classifications or self/proxy reports of dementia (k = 8; HR = 1.183, 95% CI [1.155,1.211]). (3) The association was stronger in published studies (k = 16; HR = 1.416, 95% CI [1.224,1.638]) compared to the individual cohort studies (k = 5; HR = 1.180, 95% CI [1.152,1.209]), which predominantly used performance-based classifications of dementia and ordinal/continuous scales of loneliness.

Table 2.

Meta-regressions

	All-cause Dementia

Moderators	HR	95% CI	p-value

Classification of Cognitive Status	0.764	(0.722, 0.808)	<.001
Follow-up length	0.943	(0.764, 1.164)	.586
Data Source	0.815	(0.748, 0.887)	<.001
Year of publication	0.855	(0.691, 1.059)	.151
Country
North America/US	0.894	(0.727, 1.099)	.286
Europe/UK	1.133	(0.961, 1.336)	.138
Asia/Oceania	0.945	(0.763, 1.169)	.601
Sample Size (< 3,000)	1.194	(0.995, 1.434)	.057
Incident Cases (< 10%)	1.086	(0.907, 1.300)	.371
Age (< 60 years)	0.914	(0.764, 1.092)	.321
Sex (> 60% female individuals)	0.939	(0.749, 1.176)	.582
Loneliness: UCLA scale	0.888	(0.741, 1.064)	.198
Loneliness: categorical/binary scale	1.188	(1.058, 1.335)	.004
Quality Score	1.070	(0.847, 1.352)	.572

	Cognitive Impairment
Moderators	HR	95% CI	p-value

Classification of Cognitive Status	0.900	(0.808, 1.002)	.055
Follow-up length	0.969	(0.899, 1.044)	.410
Data Source	0.920	(0.842, 1.005)	.065
Year of publication	0.939	(0.801, 1.101)	.439
Country
North America/US	0.955	(0.888, 1.026)	.205
Europe/UK	1.063	(1.005, 1.124)	.033
Asia/Oceania	0.973	(0.900, 1.052)	.493
Sample Size (< 3,000)	1.131	(1.025, 1.247)	.014
Incident Cases (< 10%)	1.027	(0.955, 1.104)	.474
Age (< 60 years)	0.973	(0.836, 1.131)	.718
Sex (> 60% female individuals)	0.984	(0.892, 1.086)	.752
Loneliness: UCLA scale	1.008	(0.936, 1.087)	.830
Loneliness: categorical/binary scale	0.984	(0.892, 1.086)	.752
Quality Score	1.065	(0.908, 1.249)	.439

Open in a new tab

The table reports meta-regression (or moderation) effects for all-cause dementia (k = 21 samples) and cognitive impairment (k = 16 samples). HR = Hazard Ratio; 95% CI = 95% Confidence Interval (lower and upper limits). Moderation was explored for assessment of cognitive status (classification based on cognitive performance or self/proxy reports vs. clinical diagnosis), follow-up length (above vs. below 10 years), data source (selected cohorts vs. published studies), year of publication (after vs. before 2020; data from individual cohorts are considered with articles published after 2020), country (North America/US vs. others; Europe/UK vs. others; Asia/Oceania vs. others), sample size (< 3,000), incident cases (< 10%), age of sample (whether included adults below 60 years of age at the baseline), proportion of female individuals (whether > 60%), loneliness measure (UCLA loneliness scale vs. others), type of loneliness scale (categorical/binary vs. ordinal/continuous) and quality score (fair vs. good). Significant effects are highlighted in bold.

Cause-specific dementia.

Five studies assessed risk for AD (N = 492,967)^5,7,9,10,36 and 3 studies for VaD (N = 489,467)^7,9,10 (Fig. 3). Loneliness was associated with an increased risk for both AD (HR = 1.393, 95% CI [1.290,1.504], p <.001; Q = 4.52, df = 4, p =.340; I² = .01%; tau² = .00) and VaD (HR = 1.735, 95% CI [1.483,2.029], p <.001; Q = 2.26, df = 2, p =.323; I² = 8.07%; tau² = .00). However, the association with VaD was driven by the inclusion of Sutin et al.⁹ (Extended Data Table 5), while the other studies found no association^7,10.

Secondary outcome = Cognitive Impairment

For CIND/CI, 16 samples were included in the meta-analysis (N = 103,387). The literature search identified 7 samples from 4 articles ^20,21,33,37 and 9 samples were from the individual cohort studies. There was a significant association in 11 of the 16 samples (Fig. 4); the pooled estimate indicated that feeling lonely increased risk of developing CIND/CI by 15% (HR = 1.150, 95% CI [1.113,1.189], p <.001). There was large heterogeneity across studies (Q = 38.63, df = 15, p <.001; I² = 58.80%; tau² = .00). The leave-one-out analysis indicated that the meta-analytic effect was not dependent on any single study (Extended Data Table 5).

Fig. 4: — The plot displays individual study estimates (16 samples, 103,387 participants) and the average random-effects (RE) model for risk of cognitive impairment. Effect sizes are hazard ratios (HR) with 95% confidence intervals (95% CI). Study abbreviations: H70 Study = Gothenberg H70 Birth Cohort Study; LEILA = Leipzig Longitudinal Study of the Aged; LRGSTUA = Neuroprotective Model for Healthy Longevity among Malaysian Older Adults Towards Using Ageing; Zhou (2019) included results for male and female (M / F) individuals. Selected cohorts: HRS = Health and Retirement Study; ELSA = English Longitudinal Study of Ageing; SHARE = Study of Health, Ageing, and Retirement in Europe; TILDA = The Irish Longitudinal Study on Ageing; MHAS = Mexican Health and Aging Study; KLoSA = Korean Longitudinal Study of Aging; CHARLS = China Health and Retirement Longitudinal Study; and HILDA = Household, Income and Labour Dynamics in Australia.

Sensitivity analysis.

Because Zhou et al.³⁷ reported sex-stratified results, we performed a sensitivity analysis substituting this study with non-stratified analyses of the same sample^12,38. The results were virtually identical when including Wei et al.³⁸ (HR = 1.147, 95% CI [1.110,1.184]) or Huang et al.¹² (HR = 1.149, 95% CI [1.112,1.187]). An additional analysis included Rawtear et al.³⁹ and Wang et al.⁴⁰, which used a statistical approach and classification of impairment that were less comparable to the other studies. The association between loneliness and CIND/CI was similar when including these two additional studies (HR = 1.148, 95% CI [1.112,1.186]). We further conducted three (unregistered) analyses: First, we ran the analysis including estimates from fully adjusted models, which controlled for depression, social isolation and/or other modifiable risk factors for dementia. The association was still significant in 7 out of 16 samples, with a meta-analytic effect that was attenuated in size (HR = 1.093, 95% CI [1.045,1.143], N = 81,709; Extended Data Fig. 1, panel b) compared to the main analysis. Second, we ran the analysis restricting results to samples aged 50 or older (N = 96,080); the effect size remained virtually the same (HR = 1.152, 95% CI [1.112,1.194]). Lastly, we included studies (k = 6) that assessed CI, but did not ascertain absence of dementia, or did not distinguish between mild and severe cognitive impairment. This analysis showed a similar risk estimate (HR = 1.145, 95% CI [1. 105,1.186]) to the main analysis.

Publication Bias.

The Egger’s test was 2.44 (p = .028), indicating asymmetry in the funnel plot (Extended Data Fig. 2, panel b). Yet, the bias-corrected trim and fill estimate was similar to that obtained in the main analysis (HR = 1.143, 95% CI [1.107,1.181]). Of note, the analysis included for the most part (unpublished) results from the selected cohort studies.

Meta-Regressions.

Two significant effects were found (Table 2): (1) The association was stronger for studies from Europe/UK (k = 7; HR = 1.177, 95% CI [1.142,1.214] vs. other countries, HR = 1.120, 95% CI [1.078,1.164]). (2) The association was stronger for samples with n < 3000 (k = 6; HR = 1.364, 95% CI [1.153,1.613]) compared to larger samples (HR = 1.132, 95% CI [1.100,1.166]).

Discussion

With 608,561 individuals pooled across 21 samples, this meta-analysis confirms the association between loneliness and risk of dementia observed in previous smaller-scale meta-analyses (< 65,000 individuals, k ≤ 10 samples)^15,19 and further identifies moderators of this association (e.g., method of dementia ascertainment). This association aligns with the literature that links loneliness with increased risk for other neurodegenerative diseases (e.g., Parkinson’s)⁴¹ and premature mortality⁴². The meta-analytic effect size is also comparable to that of well-established behavioral risk factors for dementia, such as sedentary behavior/physical inactivity (RR = 1.30; 95% CI [1.12,1.51], k = 18)⁴³, smoking (RR = 1.30, 95% CI [1.18,1.45], k = 27)⁴⁴, and other psychosocial factors (e.g., neuroticism, HR = 1.24, 95% CI [1.17,1.31], k = 12)⁴⁵ that are correlated with loneliness⁴⁶.

In addition to all-cause dementia, this meta-analysis is among the first to examine the association between loneliness and cause-specific dementia and cognitive impairment. Cause-specific dementia was assessed only in a few studies from the published literature (k = 5 for AD, k = 3 for VaD), with the meta-analysis suggesting a significant association for both dementia subtypes. For VaD, however, the association was driven by the inclusion of Sutin et al., a study with >450,000 participants⁹. The meta-analytic effect for CIND/CI (k = 16, N = 103,387) was smaller than what was observed for dementia (i.e., the confidence intervals for CIND/CI and dementia did not overlap). The association with CIND/CI is nonetheless consistent with other studies that link loneliness with poor cognitive performance and functional limitations^47,48 before the onset of noticeable dementia symptoms.

There was large heterogeneity in effect sizes across studies. For studies assessing dementia, the heterogeneity was due, in part, to differences in loneliness measures and ascertainment of dementia. For instance, most studies (16 out of 21) used a categorization of presence/absence of loneliness (with presence generally defined as feeling lonely at least sometimes), rather than assessing the intensity/frequency of loneliness. The meta-regression for the type of loneliness scale indicated that these studies tended to report a stronger association with dementia risk compared with studies that measured loneliness with continuous/ordinal scales. In addition, we found significant differences between studies that used a clinical diagnosis versus studies that used cognitive cut-offs or self/proxy reports of dementia, with the latter generally reporting smaller effect sizes. Although cognitive cut-offs or self/proxy reports of dementia are widely used^8,49,50, they might be more prone to misdiagnosis due to errors and variability in cognitive performance over time. These measures are, nonetheless, an important instrument to ascertain dementia and cognitive impairment in population cohort studies, as they are less costly than in-depth neuropsychological examinations. Further, most studies did not assess functional decline, which is a critical component of the diagnostic process. To this end, the association between loneliness and dementia is likely to be attenuated, rather than exaggerated when using cognitive cut-offs or self/proxy reports of dementia. For CIND/CI, the association was similarly dependent on the type of cognitive evaluation: a sensitivity analysis indicated that effect sizes tended to be smaller for studies that assessed CI (k = 6) but did not ascertain the absence of dementia. In addition, the meta-regression indicated that the association with CIND/CI was stronger in studies with a smaller sample size (n < 3,000), reflecting the tendency of small studies to produce more variable estimates (i.e., larger confidence intervals) and less replicable results^51,52. For CIND/CI, but not dementia, there was also evidence of funnel plot asymmetry which could be attributed to methodological/design heterogeneity across studies.

The heterogeneity across studies was not attributable to sample characteristics, such as the follow-up length. Loneliness-associated risk for both dementia and CIND/CI was not simply due to reverse causation (i.e., cognitive decline leading to feelings of loneliness), which could occur in studies with shorter intervals between loneliness and the outcomes. The associations were also independent of sample age, and estimates remained unchanged when restricting the analysis to samples aged 50 or older. Moreover, even though some studies reported negative consequences of loneliness for male, but not female individuals^14,37, we observed no differences in the associations based on the proportion of sexes within studies. For CIND/CI, the meta-regression indicated a stronger association for studies from Europe/UK (k = 7), as compared to studies from North America/US (k = 3) and Asia/Oceania (k = 6). Note, however, that studies from Europe included large cohorts, such as SHARE.

The meta-analytic effects for dementia and CIND/CI were attenuated, but still significant, in the analysis including models that accounted for depression, social isolation, and/or other modifiable risk factors. In particular, depression and social isolation emerge as possible mediators underlying the associations^25,53. Loneliness is closely associated with depressive symptomatology^24,54, which in turn increases the risk for cognitive decline⁵³. Loneliness, however, is not simply a symptom of depression, and this research suggests that loneliness has an independent relation with dementia when controlling for depressive symptoms⁹. Feeling lonely is also related to reduced engagement in social activities and poor social interactions. It is plausible that reduced social participation reduces cognitive stimulation, leaving individuals who experience loneliness more vulnerable to cognitive decline²⁵. Interventions that target social isolation and depressive symptoms may alleviate loneliness^55,56 and thus support cognitive health in the older population.

There may be other mechanisms implicated in the associations. For instance, lonely individuals tend to engage in unhealthy behaviors (e.g., physical inactivity), experience heightened stress and immune dysfunction^57–59, all of which may affect cognition. Loneliness is also related to cardiovascular risk⁶⁰. Yet, in the current analysis, the inclusion of control variables such as diabetes, hypertension, and obesity, had a negligible effect on the association with dementia and CIND/CI. Other modifiable risk factors for dementia (e.g., hearing function, air pollution)²⁶ were not considered in the current analysis because such variables were not available across all studies. These risk factors could have an important role in the association between loneliness and cognitive risk in later life.

Strengths, Limitations, and Future Directions

Strengths of this work include the systematic and comprehensive approach to identifying published studies and ongoing cohort studies that assessed loneliness and cognitive status over time. Other strengths include the consideration of cause-specific dementia and CIND/CI, publication bias, pre-registration, and the test of potential moderators. There are, however, limitations. First, the number of selected studies is still relatively small, particularly for AD and VaD. More research is needed to identify for whom and under which circumstances loneliness increases the risk of dementia/cognitive impairment and compare the risk for AD versus VaD and other cause-specific dementias⁹. Second, future studies should implement assessments that better distinguish mild to severe cognitive impairment not caused by dementia. For instance, CIND/CI is generally not associated with loss of functionality in daily life. Most studies, however, did not examine functional decline in combination with cognitive impairment. Furthermore, CIND/CI includes subjective complaints or concerns about memory, a key diagnostic feature for MCI²² that was not assessed in the current work. Third, there is a lack of data from low- and middle-income countries (i.e., Africa, Central/South America, or South Asia) and unreported information regarding race/ethnicity or cultural minorities, which highlights the need to examine socio-cultural differences underlying the consequences of loneliness in future work. Fourth, attention should be directed to other contextual and health risk factors that are relevant for both loneliness and cognition. For instance, hearing impairments could exacerbate feelings of loneliness and contribute to cognitive decline in aging adults⁶¹. Fifth, studies included in the meta-analysis assessed loneliness at one point in time, and most simply reported presence/absence of loneliness. With this classification, there is a substantial loss of information on the intensity/frequency of loneliness. For example, some studies identified a stronger risk for dementia associated with persistent as compared to transient loneliness states^62,63. Further work should identify long-term (cumulative) consequences of loneliness on cognitive aging and impairment. Lastly, researchers could also consider specific types of loneliness (perceived absence of close connections [emotional loneliness] versus absence of acceptable network [social loneliness])³⁰. Identifying the type and sources of loneliness⁶⁴ is important to develop effective interventions that address loneliness and support aging adults and their cognitive health.

Conclusions

The present work confirms the association between loneliness and risk of all-cause dementia and cause-specific dementias (AD and VaD) and extends the association to CIND/CI. The large heterogeneity across study estimates points to the need to better identify moderators that explain differences across studies. Identifying such moderators, as well as testing for mediators of the observed associations, will help identify the loneliness-associated processes that increase vulnerability to dementia.

Methods

Aging Cohort Studies

We performed a coordinated analysis of eight longitudinal cohort studies that included measures of loneliness at baseline and repeated assessments of cognitive status over time: HRS, ELSA, SHARE, TILDA, MHAS, KLoSA, CHARLS, and HILDA. Details on each study design and measures are provided in Supplementary Note S1. For each study, data were collected with the informed consent of participants; all procedures, materials, and compensation to participants were approved by the institutional review boards of their respective institutions. Please note that prior work assessed loneliness and risk of dementia and CIND/CI in HRS⁸, ELSA⁴⁹, and SHARE⁶. Since these publications, new waves of data have been released that allowed us to re-analyze the data using a longer follow-up, thereby increasing statistical power and reducing the risk of reverse causation. Further, researchers have examined loneliness and cognitive performance in TILDA^53,65 and CHARLS^66,67, but not its association with incident dementia or CIND/CI.

Loneliness.

HRS, ELSA, TILDA, and SHARE (starting at Wave 4) included the 3-item version of the well-validated UCLA Loneliness Scale⁶⁸. Respondents were asked how often they ‘lack companionship’, ‘felt left out’, and ‘isolated from others’ on a 3-point scale, from often to hardly ever or never. Items were reverse-scored and the mean taken as a measure of loneliness. This scale has good internal consistency (α = 72)⁶⁸ and has been widely used in large survey studies⁶⁹. SHARE also included a single item at Wave 1, ‘How often do you feel lonely?’ rated from almost all the time to almost none of the time (4-point scale). SHARE was analyzed using both measures of loneliness, the single-item measure with Wave 1 as baseline and the 3-item scale with Wave 4 as baseline. A similar item, ‘In the past week… I felt lonely’ was asked as part of the Center for Epidemiological Studies Depression (CES-D)^70–72 in KLoSA, CHARLS, and MHAS; response options were on a 4-point scale from rarely (or less than one day) to almost all the time (or 5–7 days), except for MHAS which used the response option yes or no. HILDA assessed loneliness with the item, ‘I often feel very lonely’, with participants rating their agreement to this statement on a 7-point scale. Single-item scores have relatively high correlations with multi-item loneliness scales (r = .73)⁶ and have shown sufficient power to predict incident dementia⁷.

Cognitive Status.

For the assessment of cognitive status, we used the classification that has been validated previously in each specific study. In the HRS, cognitive status was assessed with the modified Telephone Interview for Cognitive Status (TICSm)⁷³. The TICSm is the sum of performance on three cognitive tasks: a 10-word immediate and delayed recall memory task (20 points), serial subtractions (5 points), and backward counting (2 points). The total score is used to classify dementia (TICSm scores ≤6 out of 27) and CIND (scores between 7 and 11)⁸. Consistent with the previous publication on loneliness and dementia in ELSA⁴⁹, dementia was defined as a self-reported doctor diagnosis of Alzheimer’s disease or dementia or a score ≥ 3.5 on a short-form of the Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE), as reported by a knowledgeable proxy⁷⁴. ELSA also administered the TICSm from Wave 7 on. Using the most recent data (last four assessments), we assessed risk of developing CIND, excluding ELSA participants who reported a diagnosis of dementia or an IQCODE above the 3.5 cut-off at the baseline assessment. In TILDA and KLoSA, cognitive status was assessed with the Mini-Mental State Examination (MMSE)⁷⁵. The MMSE measured cognitive function in several domains, including orientation, attention, memory, language, and visual-spatial skills. Scores between 24 and 30 were classified as normal cognition, scores between 18 and 23 were classified as CIND, and scores < 18 were classified as dementia^76,77. For MHAS, we adopted the classification of dementia, CIND and normal function provided by the study team⁷⁸: Dementia consisted of cases with impairment in two or more cognitive domains and one or more limitations in instrumental activities of daily living (IADLs), or proxy respondents with IQCODE scores ≥ 3.4; CIND consisted of cases with impairment in two or more cognitive domains and no IADL functional limitations; unimpaired cases were respondents with normal cognitive function and no IADL limitations or impairment in IADLs due to physical limitations, and proxy respondents with IQCODE score < 3.4. For SHARE, CI was classified as in previous publications^6,79. Participants completed a 10-word memory recall and verbal fluency task; those who scored equal or below 1.5 SD of the age-graded mean of the sample on both tasks were classified as impaired. Starting from 2011 (wave 4), SHARE also recorded doctor diagnosis of Alzheimer’s disease, dementia, or senility, as reported by the participant and/or their proxy respondent; this information was used to estimate risk of dementia, as done in prior studies^50,80,81. For HILDA, CI was defined as a score equal to or below 1.5 SD of the age-graded mean of the sample on the symbol digit modality test and backward digit span^82,83. For CHARLS, a composite cognitive score was derived from a 10-word memory recall, serial subtraction, time orientation, and a figure drawing task⁸⁴; participants with a composite score equal to or below 1.5 SD of the age-graded mean of the sample were classified as impaired.

Control variables.

Sociodemographic variables were age (in years), sex (coded 1 = female individuals), race and/or ethnicity (when possible), and education. In HRS, race and ethnicity were coded as African American and other versus white and Hispanic/Latinx versus non-Hispanic/Latinx. In ELSA, race was coded as white versus not white. In HILDA, ethnicity was coded as Aboriginal or Torres Strait Islander origin versus no indigenous origin. The other studies did not report race/ethnicity. Education was reported in years in HRS and MHAS; on a scale from 0 (no qualification) to 7 (degree) in ELSA; from 1 (none/some primary) to 7 (postgraduate/higher degree) in TILDA; from 1 (<11 years) to 7 (postgraduate; masters or doctorate) in HILDA; from 1 (elementary school or lower) to 4 (college university or higher) in KLoSA and CHARLS. SHARE used the 1997 International Standard Classification of Education to categorize and harmonize education across European countries; the level of education ranged from 0 (pre-primary level of education) to 6 (second stage of tertiary education). Additional control variables were depressive symptoms, social isolation, and established risk factors for dementia: diabetes, hypertension, and obesity²⁶. Across studies, depression was measured with a version of the CES-D^72,85,86; a score was computed to reflect the presence of depressive symptoms, excluding the CES-D item that assessed loneliness. Exceptions were HILDA, which assessed depressive symptoms with the Mental Health Inventory, a subscale from the 36-item Short Form Health Survey⁸⁷, and SHARE, which assessed depressive symptoms with the EURO-D⁸⁸ (see Supplementary Table S4 for details). Social isolation was identified through the following variables: being single, separated from spouse, divorced, or widowed (1 = yes; 0 = no), having no living children (1 = yes; 0 = no), and living alone (1 = yes; 0 = no). Modifiable risk factors were doctor diagnosis of diabetes (1 = yes; 0 = no), hypertension or high blood pressure (1 = yes; 0 = no), and obesity (1=body mass index [BMI] ≥ 30; 0=BMI<30). These control variables were selected based on the availability of data from the selected cohort studies and the published literature.

Statistical Analysis.

Cox regression was used to test the relation between loneliness and risk (or hazard ratio) of developing all-cause dementia (primary outcome) and/or CIND/CI (secondary outcome) in each sample. Time was coded as time-to-incidence from baseline to the first instance of outcome over the follow-up. Participants who did not develop dementia (or CIND/CI) were censored at their last available assessment. In each sample (except MHAS), loneliness scores were standardized so that one unit corresponded to a 1 SD difference. In Model 1, loneliness was used as a predictor of each outcome, controlling for age, sex, education, race and ethnicity (where applicable). To ensure no violation of the proportional hazard assumption, we tested the interaction between time and the control variables in the model. In the case of a violation, we included the relevant interaction terms in the final model⁸⁹. Supplemental analysis excluded individuals younger than 50 years because they were less likely to develop dementia or cognitive impairment over the follow-up. In Model 2, depressive symptoms, social isolation, and modifiable factors (diabetes, hypertension, obesity) were included as additional control variables. Additional analyses controlled for depression and isolation separately from modifiable risk factors (see Extended Data Table 1).

Meta-analysis

A meta-analysis combined the results of the coordinated analysis with published studies on loneliness and risk of dementia and/or CIND/CI. The literature search and meta-analytic synthesis followed the MOOSE recommendations for meta-analyses of observational studies (see Supplementary Table S5, which reports a checklist of MOOSE recommendations). Please note the total N for each analysis varied based on the selected studies assessing all-cause dementia (N = 608,561), AD (N = 492,967), VaD (N = 489,467), and CIND/CI (N = 103,387).

Literature Search Strategy.

Five electronic databases were used: PubMed, Web of Science, and psycINFO and psycARTICLES provided by APA PsycNet, and Science Direct. Open access archives (medRxiv and PsyArXiv) were also searched to identify pre-prints of relevant studies. The searches included English publications across all years up to September 2023; during the review process, the search was updated to March 2024. The search terms were “lonel*” (to capture loneliness and lonely) OR “perceived isolation” AND “dementia” OR “Alzheimer*” OR “cognitive impairment”.

Selection Criteria and Quality Assessment.

Studies were included if the following criteria were met: (1) Study design: Longitudinal, observational studies. (2) Study population: Middle-aged and older adults at baseline who did not have dementia at the baseline assessment of loneliness, non-clinical samples. (3) Assessment of dementia and/or cognitive impairment: Studies that reported cognitive status at baseline and follow-up, through clinical diagnosis, medical or death reports, neuropsychological evaluation, self/proxy reports of a doctor diagnosis, or cut-off for cognitive performance. (4) Assessment of loneliness: Studies that measured loneliness, whether through a single item or a multi-item scale, and whether through direct questions (e.g., ‘Do you feel lonely?’) or validated scales (e.g., UCLA Loneliness Scale). A first researcher (ML) screened titles, abstracts, and keywords of each article for eligibility. If a study appeared to be eligible for the analysis, the full-text was examined. The reference lists of selected articles were also screened to identify additional studies. The full text-articles were then independently assessed for inclusion by a second researcher (DA). Both researchers (ML and DA) used the NIH quality assessment tool for observational cohort studies to rank studies by quality (Good, Fair, or Poor; see Extended Data Table 6). Discrepancies were discussed and resolved with a third author (ARS). Data Extraction File is accessible at Open Science Framework: https://osf.io/tfphs/.

Statistical Analysis.

Using the R “metafor” package⁹⁰, results across studies were combined to estimate the overall risk associated with loneliness for dementia, distinguishing all-cause dementia from AD and VaD, and CIND/CI. To reduce variability across studies, we generally chose the risk estimates from the main model, with age, sex, education, and race/ethnicity as control variables; this differs from previous meta-analyses^15,19 that included data from statistical models that accounted for the most variables, including potential mediators or colliders of the association. Follow-up analyses were conducted with results from studies that accounted for depressive symptoms, measures of social isolation, and/or modifiable risk factors for dementia. When multiple studies analyzed data from the same cohort, we generally prioritized the study with a longer follow-up to exclude the possibility of reverse causation. The pooled mean effect sizes were calculated with random-effect meta-analysis. Between-study heterogeneity was identified using Hedges Q, tau², and I² statistics. I² ranges between 0% and 100% and values exceeding 50% were considered to represent large heterogeneity. Leave-one-out analyses examined whether heterogeneity was driven by one specific study. Effects reported as odds ratios or relative risk ratios were directly considered as hazard ratios. When needed, the estimates were adjusted to reflect a 1 SD increase in loneliness using sample means and standard deviations; for one study²⁰, we further inverted the direction of estimates and confidence intervals (see Data Extraction File for details). The hazard ratios were log-transformed, and their standard errors were computed from the reported confidence intervals to conduct the analyses using “metafor”.

To identify sources of heterogeneity, we performed a series of meta-regressions: (1) We tested whether results differed between studies that used a clinical diagnosis with studies that exclusively used performance-based cut-offs or self/proxy reports of a diagnosis. (2) We tested whether the associations were moderated by follow-up length. With a short-term follow-up, there may be potential for reverse causation (i.e., cognitive decline contributing to feelings of loneliness). (3) We tested whether results varied by the location of the samples (North America/US vs. others, Europe vs. others, and Asia/Oceania vs. others). (4) We tested study characteristics and measures as moderators: quality score, data source (aging cohort vs. published study), year of publication (after vs. before 2020), sample size (>3,000), percentage of incident cases (< 10%), age of the sample (whether the sample included adults below 60 years of age), proportion of female individuals (whether more than 60%), and loneliness measures. For the latter, we grouped measures based on the type of scale used to assess loneliness (UCLA Loneliness Scale vs. single items or other scales) and whether loneliness was entered as a categorical/binary predictor (presence/absence) or as a continuous/ordinal scale in the analysis. The moderators were dichotomized to facilitate meta-regressions. Most studies did not include information on race or ethnicity, and therefore we were not able to assess the proportion of individuals from race/ethnic minorities as a moderator. We were also not able to categorize information on education because of differences in the education system and assessment/scale used across studies. To test the robustness of the results, sensitivity analyses were performed restricting the meta-analysis to samples 50 and older and including results from fully adjusted models (i.e., results controlling for depression, social isolation, and/or modifiable clinical risk factors for dementia).

We assessed potential publication bias by inspecting funnel plot asymmetry and Egger’s test and applied a trim and fill procedure to obtain a bias-corrected estimate of the overall effect. In supplemental analyses (see Supplementary Note S2), we also used the precision-effect test and precision-effect estimate with standard errors (PET-PEESE)⁹¹ to estimate the effect size that would be expected in a hypothetical study with a standard error of zero. PET adjusts for small-study effects (i.e., the tendency of studies with smaller samples to produce larger effect sizes) and assesses if there is a true effect beyond publication bias. If PET is significant (i.e., an indication of true effect), the additional PEESE provides a better estimate as it corrects for a non-linear association between the reported effect size and standard errors. We further conducted p-curve analyses^92,93 to explore the consistency and distribution of statistically significant results to assess whether or not significant findings are the result of selective reporting. The results of these analyses are reported in Supplementary Note S2.

Extended Data

Extended Data Table. 1.

Supplemental analyses from the individual cohort studies

Outcome=Dementia	Model 1 restricting analysis to participants 50 and older		Model 1 additionally controlling for depression and isolation		Model 1 additionally controlling for modifiable clinical risk factors

	HR	95% CI	HR	95% CI	HR	95% CI

HRS	1.174^***	(1.115, 1.236)	1.100^***	(1.038, 1.166)	1.184^***	(1.123, 1.248)
ELSA	1.181^***	(1.081, 1.290)	1.213^***	(1.097, 1.341)	1.155^***	(1.049, 1.272)
SHARE	1.198^***	(1.152, 1.247)	1.100^***	(1.045, 1.158)	1.199^***	(1.151, 1.249)
MHAS	1.195^**	(1.062, 1.343)	0.976	(0.847, 1.125)	1.199^*	(1.041, 1.382)
KLoSA	1.162^***	(1.112, 1.214)	1.053	(0.993, 1.117)	1.167^***	(1.118, 1.217)

Outcome = Cognitive Impairment	Model 1 restricting analysis to participants 50 and older		Model 1 additionally controlling for depression and isolation		Model 1 additionally controlling for modifiable clinical risk factors

	HR	95% CI	HR	95% CI	HR	95% CI

HRS	1.072^***	(1.038, 1.108)	1.011	(0.975, 1.049)	1.074^***	(1.039, 1.110)
ELSA	1.178^***	(1.105, 1.255)	1.141^***	(1.061, 1.226)	1.174^***	(1.098, 1.255)
SHARE Sample 1	1.148^***	(1.086, 1.214)	1.091^*	(1.013, 1.174)	1.129^***	(1.067, 1.195)
SHARE Sample 2	1.182^***	(1.123, 1.243)	1.082^*	(1.010, 1.158)	1.166^***	(1.105, 1.230)
TILDA	--		1.182^*	(1.010, 1.384)	1.310^***	(1.140, 1.505)
MHAS	1.047	(0.959, 1.142)	1.051	(0.949, 1.165)	1.020	(0.922, 1.128)
KLoSA	1.139^***	(1.099, 1.182)	1.092^***	(1.044, 1.143)	1.135^***	(1.096, 1.175)
CHARLS	1.108^***	(1.046, 1.173)	1.014	(0.944, 1.089)	1.106^***	(1.049, 1.167)
HILDA	1.357	(0.925, 1.992)	1.061	(0.724, 1.554)	1.477	(1.000, 2.183)

Open in a new tab

The table reports additional Cox regression results within each cohort. Two-tailed p-values < .05 were used for all analyses with no adjustment for multiple testing. HR = Hazard Ratio; 95% CI = 95% Confidence Interval (lower and upper limits).

^***

p ≤.001

^**

p <.01

p <.05.

Extended Data Table. 2.

Breakdown of the literature searches

PubMed	*Initial Search: 662 records* 41 Extracted 23 Deleted (see Supplemantary Information for reasons of exclusions) 18 Full text selected Search Query 1: (lonel[Title/Abstract] OR perceived isolation[Title/Abstract]) AND (dementia[Title/Abstract] OR Alzheimer[Title/Abstract] OR cognitive impairment[Title/Abstract]) Filters: English Sort by: Most Recent. No filter based on Article Type *Updated Search (2023 – 2024): 158 records* 8 Extracted 6 Duplicates (from prior search in PubMed) 1 Deleted (pre-print) 1 Full text selected Search Query 2: (lonel[Title/Abstract] OR perceived isolation[Title/Abstract]) AND (dementia[Title/Abstract] OR Alzheimer[Title/Abstract] OR cognitive impairment[Title/Abstract]) Filters: English, from 2023 – 2024 Sort by Most Recent
Web of Science	*Initial Search: 80 records* 22 Extracted 22 Duplicates (overlapping with PubMed search) 0 Deleted 0 Full text selected Search Query 1: lonel* OR perceived isolation (Title) and dementia OR Alzheimer* OR cognitive impairment (Title) and Article OR Early Access OR Review (Document Type) and Englsh (Language). Editions - A&HCI, BKCI-SSH, BKCI-S, CCR-EXPANDED, ESCI, IC, CPCI-SSH, CPCI-S, SCI-EXPANDED, SSCI. Timespan = All years. Sort by Date: newest first. *Updated Search (2023 – 2024): 17 records* 2 Extracted 2 Duplicates 0 Deleted 0 Full text selected Search Query 2: lonel* OR perceived isolation (Title) AND dementia OR Alzheimer* OR cognitive impairment (Title) AND Article OR Early Access OR Review (Document Type) AND English (Language) AND 2023 or 2024 (Publication Years). Date: newest first.
APA PsycNet	*Initial Search: 122 records* 18 Extracted 16 Duplcates (overlapping with PubMed and Web of Science searches) 2 Deleted 0 Full text selected Search Query 1: Keywords: dementia OR Alzheimer* OR cognitive impairment AND Keywords: lonel* OR perceived isolation AND Language English AND Peer-Reviewed Journals only. *Updated Search (2023 – 2024): 17 records* 1 Extracted 1 Duplicates 0 Deleted 0 Full text selected Search Query 2: Keywords: dementia OR Keywords: Alzheimer* OR Keywords: cognitive impairment AND Keywords: lonel* OR Keywords: perceived isolation AND Language: Englsh AND Peer-Reviewed Journals only AND Year: 2023 To 2024
Science Direct	*Initial Search: 171 records* 4 Extracted 4 Duplicates (overlappng with PubMed, WoS and APA PsycNet searches) 0 Deleted 0 Full text selected Search Query 1: Title, abstract, keywords: (“loneliness” OR “lonely” OR “'perceived isolation”) AND (“dementia” OR “Alzheimer” OR “cognitive impairment”); sorted by date. No filter based on Article Type *Updated Search (2023 – 2024): 42 records* 2 Extracted 2 Duplicates 0 Deleted 0 Full text selected Search Query 2: Year 2023-2024 Title, abstract, keywords: (“loneliness” OR “lonely” OR “perceived isolation”) AND (“dementia” OR “Alzheimer” OR “cognitive impairment”); sorted to date. No filter based on Article Type.
Screening prior reviews	4 Extracted 0 Duplicates 1 Deleted 3 Full text selected
medRxiv and PsyArXiv	3 Extracted 3 Duplicates (results were published in a Journal and/or identified in PubMed) 0 Deleted 0 Full text selected

Open in a new tab

Initial searches within each database were performed between May 2022 and September 2023. An updated search was conducted on March 18, 2024. A total of 49 unique full-text articles were extracted and screened for eligibility. Final decision and reasons for exclusion on each article is provided in Supplementary Information (Table S6).

Extended Data Table. 3.

Summary of the selected articles

1^st Author	Year	Study Name	Country	Loneliness Measure		Outccme of Interest	Cognitive Status Classification	Sample Size	Age M (SD), andtar range	Incident Cases (%)	Follow-up (max)	Increased Risk
1^st Author	Year	Study Name	Country	Description	Type of Scale
Dabiri	2021	Rush MAP	US	Modfied De Jong-Gierveld scale (5 items)	ordinal/continuous	MCI	clinial (post-mortem) diagnosis	747	83 (6), 50+	194 (26)	24	Yes
Mahailngam	2023	H70 Study	Sweden	single item, response scale recoded	categorical/binary	Dem, CIND/MCI	clinical diagnosis (dementia); performance-based cut-off (cognitive impaiment)	1,221	75 (6), 70+	• see date extraction.	11	No
		LCLA	Germany	single item (CES-D), response scale recoded	categorical/binary	Dem, CIND/MCI	clinical diagnosis (dementia); performance-baspd cut-off (cognitive impaiment)	1,263	82 (5), 75+	• see data extraction	18	No
		LRGSTUA	Malaysia	3-item UCLA-R response scale recoded	categorical/binary	Dem, CIND/MCI	performance-based cut-off	2,322	69 (6), 60+	• see data extraction	5	No
Sutin	2023	UK Biobank	UK	single item, response yes/no	categorical/binary	Dem, CIND/MCI	clinical diagnosis (ICD codes from hospital and death records)	492,332	57 (8)	7,475 (2)	16	Yes
Freak-Poll	2022	RS	The Netherlands	single item (CES-D), response scale recoded	categorical/binary	Dem	clinical diagnosis	4,509	71 (7), 55+	504 (11)	14	Yes
Freak-Poll	2022	SNAC-K	Sweden	single item response yes/no	categorical/binary	Dem	clinical diagnosis	2,037	72 (10), 60+	292 (14)	10	Yes
Joyce	2022	ALSOP	Australia	single item (CES-D), response scale receded	categorical/binary	Dem	clinical diagnosis	11,498	75 (4), 70–94	229 (2)	7	No
Salinas	2022	FS	US	single item (CES-D), response scale receded	categorical/binary	Dem, AD, VaD	clinical diagnosis	2,308	73 (9), 60+	329 (14)	10	Yes
Goldberg	2021	NMAP	US	Quality of Life scales, including loneliness (6 items), recoded in high vs low score	categorical/binary	Dem	performance-based cut-off	636	75 (6), 65+	100 (16)	6	No
Shibata	2021	Hisayama Study	Japan	De Jong Gierveld Loneliness Scale (6 items) recoded presence vs. absence	categorical/binary	Dem	clinical diagnosis	1,141	74 (6), 65–92	114 (10)	5	Yes
Sundström	2020	Betula Study	Sweden	single item (CES-D), response yes/no	categorical/binary	Dem, AD, VaD	clinical diagnosis	1905	73, 60+	428 (22)	20	Yes
Zhcu	2019	CLHLS	China	single item, response scale recoded	categorical/binary	CI	performance-based cut-off	3,508 F 3,390 M	65+	793 (23) F 373 (11)M	3	Yes (only for M)
Zhou	2018	CLHLS	China	single item	ordinal/continuous	Dem	selfiproxy reports of a docter diagnosis	7,867	83 (11), 65–111	394 (5)	3	Yes
Holwerda	2014	AMSTEL	The Netheriands	single item, response yes/no	categorical/binary	Dem	clinical (algorithm-based) diagnosis	2,173	65–86	158 (7)	3	Yes
Chen	2011	Anhui cohort study	China	single item, response scale recoded	categorical/binary	Dem	clinical (algorithm-based) diagnosis	1,307	65+	80 (6)	75	No
Lobo	2006	ZARADCMP Project	Spain	single item (Geriatric Mental State)	categorical/binary	Dem, MCI	categorical/binary	• see data extraction	74 (9), 55+	• See data extraction.	3	Yes (only MCI)
Wilson	2007	Rush MAP	US	Modified De Joog-Gierveld scale (5 items)	ordinal/contiruous	AD	clinical diagnosis	791	821 (7), 50+	76 (10)	4	Yes

Open in a new tab

Study abbreviations: Rush MAP = Rush Memory and Aging Project; H70 Study = Gothenberg H70 Birth Cohort Study; LEILA = Leipzig Longitudinal Study of the Aged; LRGS TUA = Neuroprotective Model for Healthy Longevity among Malaysian Older Adults Towards Using Ageing; RS = Rotterdam Study; SNAC-K = Swedish National Study on Aging and Care in Kungsholmen; ALSOP = Longitudinal Study of Older Persons, part of the ASPREE (ASPirin in Reducing Events in the Elderly) Clinical Trial; FS = Framingham Study cohorts; NMAP = North Manhattan Aging Project, subsequently also referred to as the Washington Heights/Inwood Columbia Aging Project; AMSTEL = Amsterdam Study of the Elderly; ZARADEMP = Zaragoza Dementia and Depression Project. N = number of participants/incident cases; M = Mean; SD = Standard Deviation; M / F = male/female individuals. Outcomes: Dem = dementia (all-cause); AD = Alzheimer’s disease; VaD = vascular dementia; CIND = cognitive impairment, no dementia; CI = non-specific cognitive impairment; MCI = Mild Cognitive Impairment. Cognitive Status Classification: Clinical diagnosis generally refers to a consensus process based on cognitive/neuropsychological evaluation and review of medical records; the diagnosis was based on the Diagnostic and Statistical Manual of Mental Disorders (IV or III-R) or followed the guidelines of the working group of the National Institute of Neurological and Communicative Disorders and Stroke and Alzheimer’s Disease and Related Disorders Association or the Petersen’s criteria for MCI. See Extended Data Table 4 for a description of the remaining selected articles. For further details on each study see Data Extraction File: https://osf.io/tfphs/.

Extended Data Table. 4.

Summary of the selected articles (continued)

1^st Auttor	Year	Study Name	Country	Loneliness Measure		Outcome of Interest	Cognitive Status Classification	Sample Size	Age, M (SD), ander range	Incident Cases (%)	Follow-up (max)	Increased Risk
1^st Auttor	Year	Study Name	Country	Description	Type of Scale	Outcome of Interest	Cognitive Status Classification	Sample Size	Age, M (SD), ander range	Incident Cases (%)	Follow-up (max)	Increased Risk
He	2000	--	China	No details; assumed presence vs. absence	--	AD	clinical diagnose	1,203	55+	81 (7)	10	No
Bickel	1994	--	Germany	No details; assumed presence vs. absence	--	Dem	clinical diagnosis	314	74 (6), 65–92	34 (11)	8	No
Huang	2023	CLHL.S	China	single item response scale recoded	categorical/binary	CI	performance-basad cut-off	2,732	86 (6), 80+	702 (26)	17	No
Wei	2022	CLHL.S	Chna	single item response scale recoded	categorical/binary	CI	performance-based cut-off	6,629	86 (11)	1,170(18)	3	No
Rolandi	2020	InveCe.Ab	Italy	single item response yes/no	cadgorical/binary	Dem	clinical diagnosis	1,100	70–74	111 (10)	8	No
Wang	2020	CC75C	UK	single item response scale recoded	cadgorical/binary	CI	performance-based cut-off	657	66 (4), 75+	--	20	No
Rawtear	2017	SLAS	Singapore	single item response scale recoded	cadgorical/binary	CI	performance-based cut-off	1,601	65 (7)	433 (27)	8	No

Open in a new tab

Study abbreviations: CLHLS = Chinese Longitudinal Healthy Longevity Survey; InveCe.Ab = Cerebral Aging in Abbiategrasso study; CC75C = Cambridge City over-75s Cohort study; SLAS = Singapore Longitudinal Ageing Study. N = number of participants/incident cases; M = Mean; SD = Standard Deviation; M / F = male/female individuals. Outcomes: Dem = dementia (all-cause); AD = Alzheimer’s disease; VaD = vascular dementia; CIND = cognitive impairment, no dementia; CI = non-specific cognitive impairment; MCI = Mild Cognitive Impairment. Cognitive Status Classification: Clinical diagnosis generally refers to a consensus process based on cognitive/neuropsychological evaluation and review of medical records; the diagnosis was based on the Diagnostic and Statistical Manual of Mental Disorders (IV or III-R) or followed the guidelines of the working group of the National Institute of Neurological and Communicative Disorders and Stroke and Alzheimer’s Disease and Related Disorders Association or the Petersen’s criteria for MCI. For further details on each study see Data Extraction File: https://osf.io/tfphs/. Publications highlighted in grey were selected for supplementary analysis.

Extended Data Table. 5.

Leave-one-out sensitivity analyses

	All-cause Dementia
Study	HR	95% CI	Q	I²	tau²

Mahalingam et al. (2023) [H70 Study]	1.306^***	(1.196, 1.427)	117.90^***	87.78	.02
Mahalingam et al. (2023) (LEILA)	1.315^***	(1.200, 1.440)	117.63^***	88 34	.02
Mahalingam et al. (2023) (LRGSTUA)	1.302^***	(1.193, 1.421)	116.29^***	87.46	.02
Sutin et al. (2023)	1.195^***	(1.166, 1.226)	42.50^**	2.62	.00
Freak-Poli et al. (2022) (RS)	1.306^***	(1.186, 1.437)	117.20^***	89.08	.02
Freak-Poli et al. (2022) [SNAC-K]	1.295^***	(1.188, 1.412)	115.21^***	86.95	.02
Joyce et al. (2022)	1.317^***	(1.206, 1.438)	115.90^***	87.51	.02
Salinas et al. (2022)	1 299^***	(1.187, 1.421)	11662^***	87 90	.02
Goldberg et al. (2021)	1.319^***	(1.207, 1.442)	116.39^***	87.58	.02
Shibata et al. (2021)	1.299^***	(1.187, 1.422)	116.68^***	87.94	.02
Sundstrom et al. (2020)	1.299^***	(1.184, 1.425)	116.30^***	88.33	.02
Zhou et al. (2018)	1.314^***	(1.190, 1.451)	117.76^***	89.40	.03
Holwerda et al. (2014)	1.268^***	(1.182, 1.360)	102.36^***	78.63	.01
Chen et al. (2011)	1.303^***	(1.193.1.423)	117.38^***	87.71	.02
Lobo et al. (2008)	1.300^***	(1.192, 1.419)	115.67^***	87.34	.02
Bickel & Cooper (1994)	1.318^***	(1.209, 1.437)	113.93^***	87.07	.02
HRS	1.324^***	(1.202, 1.458)	111.75^***	86.21	.02
ELSA	1.321^***	(1.199, 1.457)	117.01^***	88.46	.02
SHARE	1.321^***	(1.197, 1.458)	114.61 ^***	85.71	.02
MHAS	1.319^***	(1.196, 1.455)	117.73^***	89.03	.02
KLoSA	1.325^***	(1.204.1.458)	105.57^***	84.92	.02

	Alzheimer’s Disease
Study	HR	95% CI	Q	I²	tau²

Sutin et al. (2023)	1.524^***	(1.201, 1.934)	4.30	32.43	.02
Salinas et al. (2022)	1.385^***	(1.282, 1.496)	2.18	0.00	.00
Sundstrom et al. (2020)	1.379^***	(1.274, 1.492)	3.21	0.00	.00
Wilson et al. (2007)	1.510^***	(1.255, 1.817)	3.81	27.48	.01
He et al. (2000)	1.389^***	(1.285, 1.501)	4.21	0.03	.00

	Vascular Dementia
Study	HR	95% CI	Q	I²	tau²

Sutin et al. (2023)	1.617	(0.828, 3.159)	1.46	31.36	.10
Salinas et al. (2022)	1.675^***	(1.333.2.105)	1.53	34.70	.01
Sundstrom et al. (2020)	1.788^***	(1.599, 1.999)	0.68	0.00	.00

	Cognitive Impairment
Study	HR	95% CI	Q	I²	tau²

Dabiri et al. (2024)	1.141^***	(1.106, 1.177)	34.29^**	51.16	.00
Mahalingam et al. (2023) [H70 Study]	1.149^***	(1.112, 1.187)	37.05^***	59.75	.00
Mahalingam et al. (2023) (LEILA)	1.149^***	(1.112, 1.187)	37.76^***	59.86	.00
Mahalingam et al. (2023) (LRGSTUA)	1.151^***	(1.113, 1,190)	38.42^***	60 86	,00
Zhou et al. (2019) (F)	1.155^***	(1.116, 1.195)	37.32^***	61.92	.00
Zhou et al. (2019) [M]	1.146^***	(1.110, 1.184)	36.09^***	57.69	.00
Lobo et al. (2008)	1.146^***	(1.110, 1.182)	31.73^**	56.74	.00
HRS	1.158^***	(1.123, 1.194)	28.83^*	39.16	.00
ELSA	1.149^***	(1.107, 1.192)	36.70^***	62.38	.00
SHARE Sample 1	1.156^***	(1.111.1.203)	38.51^***	65.81	,00
SHARE Sample 2	1.149^***	(1.106, 1.192)	35.52^***	61.30	.00
TILDA	1.141^***	(1.107, 1.176)	34.03^**	51.51	.00
MH AS	1.158^***	(1.119, 1.199)	36.35^***	58.96	.00
KLoSA	1.158^***	(1.112, 1,206)	38.63^***	61.69	.00
CHALS	1.159^***	(1.115, 1.205)	38.09^***	63.42	.00
HILDA	1.148^***	(1.111, 1.186)	36 88^***	58.97	.00

Open in a new tab

The table reports the random-effects meta-analytic results for each outcome of interest when leaving out a study. HR = Hazard Ratio; 95% CI = 95% Confidence Interval (lower and upper limits). Study abbreviations: H70 Study = Gothenberg H70 Birth Cohort Study; LEILA = Leipzig Longitudinal Study of the Aged; LRGSTUA = Neuroprotective Model for Healthy Longevity among Malaysian Older Adults Towards Using Ageing; RS = Rotterdam Study; SNAC-K = Swedish National Study on Aging and Care in Kungsholmen; ALSOP = Longitudinal Study of Older Persons, part of the ASPREE (ASPirin in Reducing Events in the Elderly) Clinical Trial; M / F = male/female individuals. Selected cohorts: HRS = Health and Retirement Study; ELSA = English Longitudinal Study of Ageing; SHARE =Study of Health, Ageing and Retirement in Europe; TILDA = The Irish Longitudinal Study on Ageing; MHAS = Mexican Health and Aging Study; KLoSA = Korean Longitudinal Study of Aging, CHARLS = China Health and Retirement Longitudinal Study; and HILDA = Household, Income and Labour Dynamics in Australia.

^***

p ≤.001

^**

p <.01

p <.05.

Extended Data Table. 6.

Quality Assessment

1^st Author / Study Name	Year / Analyzed	Responses to NIH Quality Assessment Tool (or Observational Cohort studies. See notes.														Quality Score
1^st Author / Study Name	Year / Analyzed	1	2	3	4	5	6	7	8	9	10	11	12	13	14	Quality Score
Dabiri	2024	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	NR	No	Yes	Good
Mahalingam	2023	Yes	Yes	Yes	Yes	No	Yes	Yes	No	Yes	No	Yes	NR	Yes	Yes	Good
Sutin	2023	Yes	Yes	Yes	Yes	Yes	Yes	Yes	NA	Yes	No	Yes	NA	Yes	Yes	Good
Freak-Poli	2022	Yes	Yes	Yes	Yes	Yes	Yes	Yes	NA	Yes	No	Yes	NR	Yes	Yes	Good
Joyce	2022	Yes	Yes	Yes	Yes	Yes	Yes	Yes	NA	Yes	No	Yes	NR	Yes	Yes	Good
Salinas	2022	Yes	Yes	Yes	Yes	No	Yes	Yes	Yes	Yes	No	Yes	NR	Yes	Yes	Good
Goldberg	2021	Yes	Yes	No	Yes	No	Yes	Yes	Yes	Yes	Yes	Yes	NA	No	No	Fair
Shibata	2021	Yes	Yes	Yes	Yes	No	Yes	Yes	Yes	Yes	No	Yes	Yes	Yes	Yes	Good
Sundström	2020	Yes	Yes	Yes	Yes	No	Yes	Yes	NA	Yes	No	Yes	NR	Yes	Yes	Good
Zhou	2019	Yes	Yes	Yes	Yes	No	Yes	No	No	Yes	No	Yes	NA	Yes	Yes	Fair
Zhou	2018	Yes	Yes	Yes	Yes	No	Yes	No	Yes	Yes	No	No	NA	Yes	Yes	Fair
Holwerda	2014	Yes	Yes	Yes	Yes	No	Yes	No	CD	Yes	No	Yes	NR	No	Yes	Fair
Chen	2011	Yes	Yes	No	Yes	Yes	Yes	Yes	CD	No	No	Yes	NA	CD	Yes	Fair
Lobo	2008	Yes	Yes	Yes	Yes	No	Yes	No	No	Yes	No	Yes	NR	Yes	Yes	Fair
Wilson	2007	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	NA	No	Yes	Good
He	2000	Yes	Yes	Yes	Yes	No	Yes	Yes	CD	No	No	Yes	NR	No	Yes	Fair
Bickel	1994	Yes	Yes	Yes	Yes	No	Yes	Yes	CD	No	No	Yes	CD	No	Yes	Fair
Huang	2023	Yes	Yes	No	Yes	No	Yes	Yes	Yes	Yes	No	Yes	NA	No	Yes	Fair
Wei	2022	Yes	Yes	Yes	Yes	No	Yes	No	No	Yes	No	Yes	NA	Yes	Yes	Fair
Rolandi	2020	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	Yes	No	Yes	NR	Yes	Yes	Good
Wang	2020	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	NA	No	Yes	Good
Rawtaer	2017	Yes	Yes	Yes	Yes	No	Yes	Yes	No	Yes	No	No	NR	No	Yes	Fair
HRS	Analyzed	Yes	Yes	Yes	Yes	NA	Yes	Yes	Yes	Yes	No	Yes	NA	Yes	Yes	Good
ELSA	Analyzed	Yes	Yes	Yes	Yes	NA	Yes	Yes	Yes	Yes	No	Yes	NA	No	Yes	Good
SHARE	Analyzed	Yes	Yes	Yes	Yes	NA	Yes	Yes	Yes	Yes	No	Yes	NA	No	Yes	Good
TILDA	Analyzed	Yes	Yes	Yes	Yes	NA	Yes	Yes	Yes	Yes	No	Yes	NA	Yes	Yes	Good
MHAS	Analyzed	Yes	Yes	Yes	Yes	NA	Yes	Yes	No	Yes	No	Yes	NA	Yes	Yes	Good
KLoSA	Analyzed	Yes	Yes	Yes	Yes	NA	Yes	Yes	Yes	Yes	No	Yes	NA	Yes	Yes	Good
CHARLS	Analyzed	Yes	Yes	Yes	Yes	NA	Yes	Yes	Yes	Yes	No	Yes	NA	Yes	Yes	Good
HILDA	Analyzed	Yes	Yes	Yes	Yes	NA	Yes	Yes	Yes	Yes	No	No	NA	No	Yes	Good

Open in a new tab

Instructions to complete NIH Quality Assessment Tool for Observational Cohort studies are at the following webpage: https://www.nhlbi.nih.gov/health-topics/study-quality-assessment-tools. List of questions: 1. Was the research question or objective in this paper clearly stated? 2. Was the study population clearly specified and defined? 3. Was the participation rate of eligible persons at least 50%? 4. Were all the subjects selected or recruited from the same or similar populations (including the same time period)? Were inclusion and exclusion criteria for being in the study prespecified and applied uniformly to all participants? 5. Was a sample size justification, power description, or variance and effect estimates provided? 6. For the analyses in this paper, were the exposure(s) of interest measured prior to the outcome(s) being measured? 7. Was the timeframe sufficient so that one could reasonably expect to see an association between exposure and outcome if it existed? 8. For exposures that can vary in amount or level, did the study examine different levels of the exposure as related to the outcome (e.g., categories of exposure, or exposure measured as continuous variable)? 9. Were the exposure measures (independent variables) clearly defined, valid, reliable, and implemented consistently across all study participants? 10. Was the exposure(s) assessed more than once over time? 11. Were the outcome measures (dependent variables) clearly defined, valid, reliable, and implemented consistently across all study participants? 12. Were the outcome assessors blinded to the exposure status of participants? 13. Was loss to follow-up after baseline 20% or less? 14. Were key potential confounding variables measured and adjusted statistically for their impact on the relationship between exposure(s) and outcome(s)? Response options: Yes/No; CD = Cannot Determine; NA = Not Applicable; NR = Not Reported. Selected cohorts: HRS = Health and Retirement Study; ELSA = English Longitudinal Study of Ageing; SHARE = Study of Health, Ageing and Retirement in Europe; TILDA = The Irish Longitudinal Study on Ageing; MHAS = Mexican Health and Aging Study; KLoSA = Korean Longitudinal Study of Aging; CHARLS = China Health and Retirement Longitudinal Study; and HILDA = Household, Income and Labour Dynamics in Australia. For further details on each study see Data Extraction File: https://osf.io/tfphs/. Publications highlighted in grey were selected for supplementary analysis.

Supplementary Material

Supplementary Information

NIHMS2029056-supplement-Supplementary_Information.pdf^{(2.2MB, pdf)}

Acknowledgments

The work reported in this publication was supported by the National Institute on Aging of the National Institutes of Health: Grant Numbers R01AG074573 and RF1AG053297 (A.R.S.), and R01AG068093 (A.T.). The funders had no role in the study design, analysis, decision to publish or preparation of the manuscript.

We further thank all participants, national and international agencies that support each cohort study analyzed in this work: HRS (Health and Retirement Study) sponsored by the US National Institute on Aging (grant number NIA U01AG009740) and coordinated by the University of Michigan; ELSA (English Longitudinal Study on Ageing) sponsored by the US National Institute on Aging (grant number NIA R01AG017644) and the UK Government Departments coordinated by the National Institute for Health and Care Research; SHARE (Survey of Health, Ageing and Retirement in Europe) funded by the European Commission and Horizon 2020, and supported by the German Ministry of Education and Research, the Max Planck Society for the Advancement of Science, the US National Institute on Aging and various national funding sources (see www.share-project.org); TILDA (Irish Longitudinal Study on Ageing) based in Trinity College Dublin for which data are hosted by the Irish Social Science Data Archive and the Interuniversity Consortium for Political and Social Research (ICPSR) based in the University of Michigan; MHAS (Mexican Health and Aging Study) supported by the US National Institute on Aging (grant number NIA R01AG018016) with the collaborative effort from the University of Texas Medical Branch (UTMB), the Instituto Nacional de Estadística y Geografía (INEGI, Mexico), the University of Wisconsin, the Instituto Nacional de Geriatría (INGer, Mexico) and the Instituto Nacional de Salud Pública (INSP, Mexico); KLoSA (Korean Longitudinal Study of Aging) for which data are cured by the Korea Employment Information Service; CHARLS (China Health and Retirement Longitudinal Study) supported by the US National Institute on Aging (grant number NIA R01AG037031), the Natural Science Foundation of China, the World Bank, and Peking University; HILDA (Household, Income and Labour Dynamics in Australia) funded by Australian Government Department of Social Services and managed by the Melbourne Institute of Applied Economic and Social Research. For each study, data were collected with the informed consent of participants; all procedures, materials and participant compensations were approved by the institutional review boards of their respective institutions.

Footnotes

Competing interests

The authors declare no competing interests.

Code availability

The R script that supports meta-analytic results is available in the Open Science Framework repository for social sciences. Access link: https://osf.io/tfphs/ (Luchetti et al. Loneliness and risk for dementia. doi: 10.17605/OSF.IO/TFPHS).

Contributor Information

Martina Luchetti, Department of Behavioral Sciences and Social Medicine, Florida State University College of Medicine, Tallahassee, FL, USA.

Damaris Aschwanden, Department of Geriatrics, Florida State University College of Medicine, Tallahassee, FL, USA. Center for the Interdisciplinary Study of Gerontology and Vulnerability, University of Geneva, Geneva, Switzerland.

Amanda A. Sesker, Department of Behavioral Sciences and Social Medicine, Florida State University College of Medicine, Tallahassee, FL, USA.

Xianghe Zhu, Department of Psychology, School of Mental Health, Institute of Aging, Key Laboratory of Alzheimer’s Disease of Zhejiang Province, and Zhejiang Provincial Clinical Research Center for Mental Disorders, The Affiliated Kangning Hospital, Wenzhou Medical University, Wenzhou, Zhejiang, China. Oujiang Laboratory (Zhejiang Lab for Regenerative Medicine, Vision and Brain Health), Wenzhou, Zhejiang, China.

Páraic S. O’Súilleabháin, Department of Psychology, University of Limerick, Limerick, Ireland. Health Research Institute, University of Limerick, Limerick, Ireland.

Yannick Stephan, Euromov, University of Montpellier, Montpellier, France.

Antonio Terracciano, Department of Geriatrics, Florida State University College of Medicine, Tallahassee, FL, USA.

Angelina R. Sutin, Department of Behavioral Sciences and Social Medicine, Florida State University College of Medicine, Tallahassee, FL, USA.

Data availability

The present study includes a coordinated analysis of data from eight public cohort studies: HRS, https://hrs.isr.umich.edu/data-products; ELSA, https://www.elsa-project.ac.uk/accessing-elsa-data; SHARE, https://share-eric.eu/data/data-access; TILDA, https://www.icpsr.umich.edu/web/ICPSR/series/726; MHAS, https://www.mhasweb.org/DataProducts/Home.aspx; KLoSA, https://survey.keis.or.kr/eng/myinfo/login.jsp; CHARLS, https://charls.charlsdata.com/pages/data/111/en.html; and HILDA, https://melbourneinstitute.unimelb.edu.au/hilda/for-data-users. Our access to the data does not allow for data redistribution. Individual researchers can access data from each of these studies after registration at each study data portal; we described each cohort in detail in Supplementary Information. Data for the meta-analysis are available in the Open Science Framework repository for social sciences. Access link: https://osf.io/tfphs/ (Luchetti et al. Loneliness and risk for dementia. doi: 10.17605/OSF.IO/TFPHS).

References

1.National Academies of Sciences, Engineering, and Medicine. Social Isolation and Loneliness in Older Adults: Opportunities for the Health Care System. (National Academies Press, Washington, D.C., 2020). [PubMed] [Google Scholar]
2.Office of the Surgeon General. Our epidemic of loneliness and isolation: The U.S. Surgeon General’s Advisory on the healing effects of social connection and community (U.S. Department of Health and Human Services, Washington, D.C., 2023). [PubMed] [Google Scholar]
3.Hawkley LC & Cacioppo JT Loneliness matters: A theoretical and empirical review of consequences and mechanisms. Ann Behav Med. 40, 218–227 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
4.Cacioppo JT et al. Loneliness within a nomological net: An evolutionary perspective. J Res Pers. 40, 1054–1085 (2006). [Google Scholar]
5.Wilson RS et al. Loneliness and risk of alzheimer disease. Arch Gen Psychiatry. 64, 234 (2007). [DOI] [PubMed] [Google Scholar]
6.Luchetti M et al. Loneliness is associated with risk of cognitive impairment in the Survey of Health, Ageing and Retirement in Europe. Int J Geriatr Psychiatry. 35, 794–801 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
7.Sundström A, Adolfsson AN, Nordin M & Adolfsson R Loneliness increases the risk of all-cause dementia and Alzheimer’s disease. J Gerontol B Psychol Sci Soc Sci. 75, 919–926 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
8.Sutin AR, Stephan Y, Luchetti M & Terracciano A Loneliness and risk of dementia. J Gerontol B Psychol Sci Soc Sci. 75, 1414–1422 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
9.Sutin AR et al. Loneliness and risk of all-cause, Alzheimer’s, vascular, and frontotemporal dementia: A prospective study of 492,322 individuals over 15 years. Int. Psychogeriatr. 35, 283–292 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Salinas J et al. Association of loneliness With 10-year dementia risk and early markers of vulnerability for neurocognitive decline. Neurology. 98, e1337–e1348 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
11.Goldberg TE, Choi J, Lee S, Gurland B & Devanand DP Effects of restriction of activities and social isolation on risk of dementia in the community. Int Psychogeriatr. 33, 1207–1215 (2021). [DOI] [PubMed] [Google Scholar]
12.Huang Y, Zhu X, Liu X & Li J The effects of loneliness, social isolation, and associated gender differences on the risk of developing cognitive impairment for Chinese oldest old. Aging Ment Health. 27, 1360–1367 (2023). [DOI] [PubMed] [Google Scholar]
13.Joyce J et al. Social isolation, social support, and loneliness and their relationship with cognitive health and dementia. Int J Geriat Psychiatry. 37, gps.5644 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
14.Zhou Z, Wang P & Fang Y Loneliness and the risk of dementia among older Chinese adults: Gender differences. Aging Ment Health. 22, 519–525 (2018). [DOI] [PubMed] [Google Scholar]
15.Lara E et al. Does loneliness contribute to mild cognitive impairment and dementia? A systematic review and meta-analysis of longitudinal studies. Ageing Res Rev. 52, 7–16 (2019). [DOI] [PubMed] [Google Scholar]
16.Lazzari C & Rabottini M COVID-19, loneliness, social isolation and risk of dementia in older people: A systematic review and meta-analysis of the relevant literature. Int J Psychiatry Clin Pract. 26, 196–207 (2022). [DOI] [PubMed] [Google Scholar]
17.Piolatto M et al. The effect of social relationships on cognitive decline in older adults: An updated systematic review and meta-analysis of longitudinal cohort studies. BMC Public Health. 22, 278 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Wang S, Molassiotis A, Guo C, Leung ISH & Leung AYM Association between social integration and risk of dementia: A systematic review and meta-analysis of longitudinal studies. J Am Geriatr Soc. 71, 632–645 (2023). [DOI] [PubMed] [Google Scholar]
19.Qiao L et al. Association between loneliness and dementia risk: A systematic review and meta-analysis of cohort studies. Front Hum Neurosci. 16, 899814 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
20.Mahalingam G et al. Social connections and risk of incident mild cognitive impairment, dementia, and mortality in 13 longitudinal cohort studies of ageing. Alzheimers Dem. 9, 5114–5128 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
21.Dabiri S, Mwendwa DT & Campbell A Psychological and neurobiological mechanisms involved in the relationship between loneliness and cognitive decline in older adults. Brain Behav Immun. 116, 10–21 (2024). [DOI] [PubMed] [Google Scholar]
22.Petersen RC et al. Practice guideline update summary: Mild cognitive impairment: Report of the Guideline Development, Dissemination, and Implementation Subcommittee of the American Academy of Neurology. Neurology. 90, 126–135 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
23.Willroth EC & Atherton OE Best laid plans: A guide to reporting preregistration deviations. Adv Methods Pract Psychol Sci. 7 (2024). [Google Scholar]
24.Luo M Social Isolation, loneliness, and depressive symptoms: A twelve-year population study of temporal dynamics. J Gerontol B Psychol Sci Soc Sci. 78, 280–290 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
25.Kim AJ, Beam CR, Greenberg NE & Burke SL Health factors as potential mediators of the longitudinal effect of loneliness on general cognitive ability. Am J Geriatr Psychiatry. 28, 1272–1283 (2020). [DOI] [PubMed] [Google Scholar]
26.Livingston G et al. Dementia prevention, intervention, and care: 2020 report of the Lancet Commission. Lancet. 396, 413–446 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Stroup DF et al. Meta-analysis of observational studies in epidemiology: A proposal for reporting. JAMA. 283, 2008–2012 (2000). [DOI] [PubMed] [Google Scholar]
28.Dubois B, Von Arnim CAF, Burnie N, Bozeat S & Cummings J Biomarkers in Alzheimer’s disease: Role in early and differential diagnosis and recognition of atypical variants. Alz Res Therapy. 15, 175 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
29.Freak-Poli R et al. Loneliness, Not social support, is associated with cognitive decline and dementia across two longitudinal population-based cohorts. J Alzheimers Dis. 85, 295–308 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
30.Shibata M et al. Emotional loneliness is associated with a risk of dementia in a general japanese older population: The Hisayama Study. J Gerontol B Psychol Sci Soc Sci. 76, 1756–1766 (2021). [DOI] [PubMed] [Google Scholar]
31.Holwerda TJ et al. Feelings of loneliness, but not social isolation, predict dementia onset: results from the Amsterdam Study of the Elderly (AMSTEL). J Neurol Neurosurg Psychiatry. 85, 135–142 (2014). [DOI] [PubMed] [Google Scholar]
32.Chen R et al. Incident dementia in a defined older chinese population. PLoS One. 6, e24817 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
33.Lobo A et al. Non-cognitive psychopathological symptoms associated with incident mild cognitive impairment and dementia, alzheimer’s type. Neurotox Res. 14, 263–272 (2008). [DOI] [PubMed] [Google Scholar]
34.Bickel H & Cooper B Incidence and relative risk of dementia in an urban elderly population: Findings of a prospective field study. Psychol Med. 24, 179–192 (1994). [DOI] [PubMed] [Google Scholar]
35.Rolandi E et al. Estimating the potential for dementia prevention through modifiable risk factors elimination in the real-world setting: A population-based study. Alz Res Therapy. 12, 94 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
36.He YL, Zhang XK, & Zhang MY Psychosocial risk factors for Alzheimer’s disease. Hong Kong J Psychiatry. 10, (2000). [Google Scholar]
37.Zhou Z et al. The association between loneliness and cognitive impairment among older men and women in China: A nationwide longitudinal study. Int J Environ Res Public Health. 16, 2877 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
38.Wei K et al. Living arrangement modifies the associations of loneliness with adverse health outcomes in older adults: Evidence from the CLHLS. BMC Geriatr. 22, 59 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
39.Rawtaer I et al. Psychosocial risk and protective factors and incident mild cognitive impairment and dementia in community dwelling elderly: Findings from the Singapore Longitudinal Ageing Study. J Alzheimers Dis. 57, 603–611 (2017). [DOI] [PubMed] [Google Scholar]
40.Wang H et al. Longitudinal analysis of the impact of loneliness on cognitive function over a 20-year follow-up. Aging Ment Health. 24, 1815–1821 (2020). [DOI] [PubMed] [Google Scholar]
41.Terracciano A, Luchetti M, Karakose S, Stephan Y & Sutin AR Loneliness and risk of Parkinson disease. JAMA Neurol. 80, 1138 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
42.Wang F et al. A systematic review and meta-analysis of 90 cohort studies of social isolation, loneliness, and mortality. Nat Hum Behav. 7, 1307–1319 (2023). [DOI] [PubMed] [Google Scholar]
43.Yan S et al. Association between sedentary behavior and the risk of dementia: A systematic review and meta-analysis. Transl Psychiatry. 10, 112 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
44.Zhong G, Wang Y, Zhang Y, Guo JJ & Zhao Y Smoking is associated with an increased risk of dementia: A meta-analysis of prospective cohort studies with investigation of potential effect modifiers. PLoS One. 10, e0118333 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
45.Aschwanden D et al. Is personality associated with dementia risk? A meta-analytic investigation. Ageing Res Rev. 67, (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
46.Buecker S, Maes M, Denissen JJA & Luhmann M Loneliness and the Big Five personality traits: A meta–analysis. Eur J Pers. 34, 8–28 (2020). [Google Scholar]
47.Perissinotto CM, Stijacic Cenzer I & Covinsky KE Loneliness in older persons: A predictor of functional decline and death. Arch Intern Med. 172, (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
48.Cachón-Alonso L, Hakulinen C, Jokela M, Komulainen K & Elovainio M Loneliness and cognitive function in older adults: Longitudinal analysis in 15 countries. Psychol Aging. 38, 778–789 (2023). [DOI] [PubMed] [Google Scholar]
49.Rafnsson SB, Orrell M, d’Orsi E, Hogervorst E & Steptoe A Loneliness, social integration, and incident dementia over 6 years: Prospective findings from the English Longitudinal Study of Ageing. J Gerontol B Psychol Sci Soc Sci. 75, 114–124 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
50.Ferreira RG, Brandão MP & Cardoso MF An update of the profile of older adults with dementia in Europe: Findings from SHARE. Aging Ment Health. 24, 374–381 (2020). [DOI] [PubMed] [Google Scholar]
51.Button KS et al. Power failure: Why small sample size undermines the reliability of neuroscience. Nat Rev Neurosci. 14, 365–376 (2013). [DOI] [PubMed] [Google Scholar]
52.IntHout J, Ioannidis JPA, Borm GF & Goeman JJ Small studies are more heterogeneous than large ones: A meta-meta-analysis. J Clin Epidemiol. 68, 860–869 (2015). [DOI] [PubMed] [Google Scholar]
53.McHugh Power J, Tang J, Kenny RA, Lawlor BA & Kee F Mediating the relationship between loneliness and cognitive function: The role of depressive and anxiety symptoms. Aging Ment Health. 24, 1071–1078 (2020). [DOI] [PubMed] [Google Scholar]
54.Cacioppo JT, Hawkley LC & Thisted RA Perceived social isolation makes me sad: 5-year cross-lagged analyses of loneliness and depressive symptomatology in the Chicago Health, Aging, and Social Relations Study. Psychol Aging. 25, 453–463 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
55.O’Rourke HM, Collins L & Sidani S Interventions to address social connectedness and loneliness for older adults: A scoping review. BMC Geriatr. 18, 214 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
56.Joshi P et al. Social connections as determinants of cognitive health and as targets for social interventions in persons with or at risk of Alzheimer’s disease and related disorders: a scoping review. Int Psychogeriatr. 36, 92–118 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
57.Hawkley LC, Thisted RA & Cacioppo JT Loneliness predicts reduced physical activity: Cross-sectional & longitudinal analyses. Health Psychol. 28, 354–363 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
58.Christiansen J, Larsen FB & Lasgaard M Do stress, health behavior, and sleep mediate the association between loneliness and adverse health conditions among older people? Soc Sci & Med. 152, 80–86 (2016). [DOI] [PubMed] [Google Scholar]
59.Shiovitz-Ezra S & Parag O Does loneliness ‘get under the skin’? Associations of loneliness with subsequent change in inflammatory and metabolic markers. Aging Ment Health. 23, 1358–1366 (2019). [DOI] [PubMed] [Google Scholar]
60.Paul E, Bu F & Fancourt D Loneliness and risk for cardiovascular disease: Mechanisms and future directions. Curr Cardiol Rep. 23, 68 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
61.Maharani A, Pendleton N & Leroi I Hearing Impairment, loneliness, social isolation, and cognitive function: Longitudinal analysis using English Longitudinal Study on Ageing. Am J Geriatr Psychiatry. 27, 1348–1356 (2019). [DOI] [PubMed] [Google Scholar]
62.Akhter-Khan SC et al. Associations of loneliness with risk of Alzheimer’s disease dementia in the Framingham Heart Study. Alzheimer’s & Dementia 17, 1619–1627 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
63.Li Y et al. Eight-year trajectories of late-life loneliness and incident dementia: A nationally representative cohort study. Am J Geriatr Psychiatry. 31, 475–486 (2023). [DOI] [PubMed] [Google Scholar]
64.Akhter-Khan SC, Prina M, Wong GH-Y, Mayston R & Li L Understanding and addressing older adults’ loneliness: The Social Relationship Expectations Framework. Perspect Psychol Sci. 18, 762–777 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
65.McHugh Power JE et al. Lonely SARTs: Loneliness and sustained attention in the Irish Longitudinal Study of Aging. Neuropsychol Dev Cogn B Aging Neuropsychol Cogn. 27, 197–206 (2020). [DOI] [PubMed] [Google Scholar]
66.Yang R et al. Loneliness as a mediator of the impact of social isolation on cognitive functioning of Chinese older adults. Age Ageing. 49, 599–604 (2020). [DOI] [PubMed] [Google Scholar]
67.Yu B, Steptoe A, Chen Y & Jia X Social isolation, rather than loneliness, is associated with cognitive decline in older adults: The China Health and Retirement Longitudinal Study. Psychol Med. 51, 2414–2421 (2021). [DOI] [PubMed] [Google Scholar]
68.Hughes ME, Waite LJ, Hawkley LC & Cacioppo JT A short scale for measuring loneliness in large surveys: Results from two population-based studies. Res Aging. 26, 655–672 (2004). [DOI] [PMC free article] [PubMed] [Google Scholar]
69.Smith J, Ryan LH, Fisher GG, Sonnega A, & Weir DR HRS Psychosocial and Lifestyle Questionnaire 2006–2016. (Survey Research Center, Institute for Social Research, Michigan, 2017). [Google Scholar]
70.Radloff LS The CES-D scale: A self-report depression scale for research in the general population. Appl Psychol Meas. 1, 385–401 (1977). [Google Scholar]
71.Lewinsohn PM, Seeley JR, Roberts RE & Allen NB Center for Epidemiologic Studies Depression Scale (CES-D) as a screening instrument for depression among community-residing older adults. Psychol Aging. 12, 277–287 (1997). [DOI] [PubMed] [Google Scholar]
72.Irwin M, Artin KH & Oxman MN Screening for depression in the older adult: Criterion validity of the 10-item Center for Epidemiological Studies Depression Scale (CES-D). Arch Intern Med. 159, 1701 (1999). [DOI] [PubMed] [Google Scholar]
73.Crimmins EM, Kim JK, Langa KM & Weir DR Assessment of cognition using surveys and neuropsychological assessment: The Health and Retirement Study and the Aging, Demographics, and Memory Study. J Gerontol B Psychol Sci Soc Sci. 66B, i162–i171 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
74.Quinn TJ et al. Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE) for the diagnosis of dementia within community dwelling populations. Cochrane Database Syst Rev. CD010079 (2014). [DOI] [PubMed] [Google Scholar]
75.Folstein MF, Folstein SE & McHugh PR “Mini-mental state”. A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res. 12, 189–198 (1975). [DOI] [PubMed] [Google Scholar]
76.Bassuk SS, Wypij D & Berkmann LF Cognitive impairment and mortality in the community-dwelling elderly. Am J Epidemiol. 151, 676–688 (2000). [DOI] [PubMed] [Google Scholar]
77.Creavin ST et al. Mini-Mental State Examination (MMSE) for the detection of Alzheimer’s dementia and other dementias in asymptomatic and previously clinically unevaluated people aged over 65 years in community and primary care populations. Cochrane Database Syst Rev. CD011145 (2016).. [DOI] [PMC free article] [PubMed] [Google Scholar]
78.Mejia-Arango S et al. Effect of demographic and health dynamics on cognitive status in Mexico between 2001 and 2015: Evidence from the Mexican Health and Aging Study. Geriatrics. 6, 63 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
79.Lugo-Palacios DG & Gannon B Health care utilisation amongst older adults with sensory and cognitive impairments in Europe. Health Econ Rev. 7, 44 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
80.Giné-Garriga M et al. Is loneliness a predictor of the modern geriatric giants? Analysis from the Survey of Health, Ageing, and Retirement in Europe. Maturitas. 144, 93–101 (2021). [DOI] [PubMed] [Google Scholar]
81.Gontié R et al. Relationship between physical activity and incidence of dementia in people aged 50 and over in Europe. Aging Ment Health. 27, 1429–1435 (2023). [DOI] [PubMed] [Google Scholar]
82.Aschwanden D, Sutin AR, Luchetti M, Stephan Y & Terracciano A Personality and dementia risk in England and Australia. GeroPsych 33, 197–208 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
83.Hosking DE & Anstey KJ The economics of cognitive impairment: Volunteering and cognitive function in the HILDA Survey. Gerontology. 62, 536–540 (2016). [DOI] [PubMed] [Google Scholar]
84.Zhang K & Zhang W Adverse childhood experiences and mild cognitive impairment in later life: Exploring rural/urban and gender differences using CHARLS. J Appl Gerontol. 41, 1454–1464 (2022). [DOI] [PubMed] [Google Scholar]
85.Karim J, Weisz R, Bibi Z & ur Rehman S Validation of the eight-item Center for Epidemiologic Studies Depression Scale (CES-D) among older adults. Curr Psychol. 34, 681–692 (2015). [Google Scholar]
86.Kim H, Park S-M, Jang S-N & Kwon S Depressive symptoms, chronic medical illness, and health care utilization: Findings from the Korean Longitudinal Study of Ageing (KLoSA). Int Psychogeriatr. 23, 1285–1293 (2011). [DOI] [PubMed] [Google Scholar]
87.Ware JE & Gandek B Overview of the SF-36 Health Survey and the International Quality of Life Assessment (IQOLA) Project. J Clin Epidemiol. 51, 903–912 (1998). [DOI] [PubMed] [Google Scholar]
88.Castro-Costa E et al. Ascertaining late-life depressive symptoms in Europe: An evaluation of the survey version of the EURO-D scale in 10 nations. The SHARE project. Int J Methods Psychiatr Res. 17, 12–29 (2008). [DOI] [PMC free article] [PubMed] [Google Scholar]
89.Tabachnick BG & Fidell LS Using multivariate statistics. (Tabachnick, Tabachnick, 2013). [Google Scholar]
90.Viechtbauer W Conducting meta-analyses in R with the metafor Package. J Stat Soft. 36, (2010). [Google Scholar]
91.Stanley TD & Doucouliagos H Meta-regression approximations to reduce publication selection bias. Res Synth Methods. 5, 60–78 (2014). [DOI] [PubMed] [Google Scholar]
92.Simonsohn U, Nelson LD & Simmons JP P-curve: A key to the file-drawer. J Exp Psychol Gen. 143, 534–547 (2014). [DOI] [PubMed] [Google Scholar]
93.Simonsohn U, Simmons JP & Nelson LD Better P-curves: Making P-curve analysis more robust to errors, fraud, and ambitious P-hacking, a reply to Ulrich and Miller (2015). J Exp Psychol Gen. 144, 1146–1152 (2015). [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Information

NIHMS2029056-supplement-Supplementary_Information.pdf^{(2.2MB, pdf)}

Data Availability Statement

[R1] 1.National Academies of Sciences, Engineering, and Medicine. Social Isolation and Loneliness in Older Adults: Opportunities for the Health Care System. (National Academies Press, Washington, D.C., 2020). [PubMed] [Google Scholar]

[R2] 2.Office of the Surgeon General. Our epidemic of loneliness and isolation: The U.S. Surgeon General’s Advisory on the healing effects of social connection and community (U.S. Department of Health and Human Services, Washington, D.C., 2023). [PubMed] [Google Scholar]

[R3] 3.Hawkley LC & Cacioppo JT Loneliness matters: A theoretical and empirical review of consequences and mechanisms. Ann Behav Med. 40, 218–227 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R4] 4.Cacioppo JT et al. Loneliness within a nomological net: An evolutionary perspective. J Res Pers. 40, 1054–1085 (2006). [Google Scholar]

[R5] 5.Wilson RS et al. Loneliness and risk of alzheimer disease. Arch Gen Psychiatry. 64, 234 (2007). [DOI] [PubMed] [Google Scholar]

[R6] 6.Luchetti M et al. Loneliness is associated with risk of cognitive impairment in the Survey of Health, Ageing and Retirement in Europe. Int J Geriatr Psychiatry. 35, 794–801 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R7] 7.Sundström A, Adolfsson AN, Nordin M & Adolfsson R Loneliness increases the risk of all-cause dementia and Alzheimer’s disease. J Gerontol B Psychol Sci Soc Sci. 75, 919–926 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R8] 8.Sutin AR, Stephan Y, Luchetti M & Terracciano A Loneliness and risk of dementia. J Gerontol B Psychol Sci Soc Sci. 75, 1414–1422 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R9] 9.Sutin AR et al. Loneliness and risk of all-cause, Alzheimer’s, vascular, and frontotemporal dementia: A prospective study of 492,322 individuals over 15 years. Int. Psychogeriatr. 35, 283–292 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R10] 10.Salinas J et al. Association of loneliness With 10-year dementia risk and early markers of vulnerability for neurocognitive decline. Neurology. 98, e1337–e1348 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R11] 11.Goldberg TE, Choi J, Lee S, Gurland B & Devanand DP Effects of restriction of activities and social isolation on risk of dementia in the community. Int Psychogeriatr. 33, 1207–1215 (2021). [DOI] [PubMed] [Google Scholar]

[R12] 12.Huang Y, Zhu X, Liu X & Li J The effects of loneliness, social isolation, and associated gender differences on the risk of developing cognitive impairment for Chinese oldest old. Aging Ment Health. 27, 1360–1367 (2023). [DOI] [PubMed] [Google Scholar]

[R13] 13.Joyce J et al. Social isolation, social support, and loneliness and their relationship with cognitive health and dementia. Int J Geriat Psychiatry. 37, gps.5644 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R14] 14.Zhou Z, Wang P & Fang Y Loneliness and the risk of dementia among older Chinese adults: Gender differences. Aging Ment Health. 22, 519–525 (2018). [DOI] [PubMed] [Google Scholar]

[R15] 15.Lara E et al. Does loneliness contribute to mild cognitive impairment and dementia? A systematic review and meta-analysis of longitudinal studies. Ageing Res Rev. 52, 7–16 (2019). [DOI] [PubMed] [Google Scholar]

[R16] 16.Lazzari C & Rabottini M COVID-19, loneliness, social isolation and risk of dementia in older people: A systematic review and meta-analysis of the relevant literature. Int J Psychiatry Clin Pract. 26, 196–207 (2022). [DOI] [PubMed] [Google Scholar]

[R17] 17.Piolatto M et al. The effect of social relationships on cognitive decline in older adults: An updated systematic review and meta-analysis of longitudinal cohort studies. BMC Public Health. 22, 278 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R18] 18.Wang S, Molassiotis A, Guo C, Leung ISH & Leung AYM Association between social integration and risk of dementia: A systematic review and meta-analysis of longitudinal studies. J Am Geriatr Soc. 71, 632–645 (2023). [DOI] [PubMed] [Google Scholar]

[R19] 19.Qiao L et al. Association between loneliness and dementia risk: A systematic review and meta-analysis of cohort studies. Front Hum Neurosci. 16, 899814 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R20] 20.Mahalingam G et al. Social connections and risk of incident mild cognitive impairment, dementia, and mortality in 13 longitudinal cohort studies of ageing. Alzheimers Dem. 9, 5114–5128 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R21] 21.Dabiri S, Mwendwa DT & Campbell A Psychological and neurobiological mechanisms involved in the relationship between loneliness and cognitive decline in older adults. Brain Behav Immun. 116, 10–21 (2024). [DOI] [PubMed] [Google Scholar]

[R22] 22.Petersen RC et al. Practice guideline update summary: Mild cognitive impairment: Report of the Guideline Development, Dissemination, and Implementation Subcommittee of the American Academy of Neurology. Neurology. 90, 126–135 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R23] 23.Willroth EC & Atherton OE Best laid plans: A guide to reporting preregistration deviations. Adv Methods Pract Psychol Sci. 7 (2024). [Google Scholar]

[R24] 24.Luo M Social Isolation, loneliness, and depressive symptoms: A twelve-year population study of temporal dynamics. J Gerontol B Psychol Sci Soc Sci. 78, 280–290 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R25] 25.Kim AJ, Beam CR, Greenberg NE & Burke SL Health factors as potential mediators of the longitudinal effect of loneliness on general cognitive ability. Am J Geriatr Psychiatry. 28, 1272–1283 (2020). [DOI] [PubMed] [Google Scholar]

[R26] 26.Livingston G et al. Dementia prevention, intervention, and care: 2020 report of the Lancet Commission. Lancet. 396, 413–446 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R27] 27.Stroup DF et al. Meta-analysis of observational studies in epidemiology: A proposal for reporting. JAMA. 283, 2008–2012 (2000). [DOI] [PubMed] [Google Scholar]

[R28] 28.Dubois B, Von Arnim CAF, Burnie N, Bozeat S & Cummings J Biomarkers in Alzheimer’s disease: Role in early and differential diagnosis and recognition of atypical variants. Alz Res Therapy. 15, 175 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R29] 29.Freak-Poli R et al. Loneliness, Not social support, is associated with cognitive decline and dementia across two longitudinal population-based cohorts. J Alzheimers Dis. 85, 295–308 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R30] 30.Shibata M et al. Emotional loneliness is associated with a risk of dementia in a general japanese older population: The Hisayama Study. J Gerontol B Psychol Sci Soc Sci. 76, 1756–1766 (2021). [DOI] [PubMed] [Google Scholar]

[R31] 31.Holwerda TJ et al. Feelings of loneliness, but not social isolation, predict dementia onset: results from the Amsterdam Study of the Elderly (AMSTEL). J Neurol Neurosurg Psychiatry. 85, 135–142 (2014). [DOI] [PubMed] [Google Scholar]

[R32] 32.Chen R et al. Incident dementia in a defined older chinese population. PLoS One. 6, e24817 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R33] 33.Lobo A et al. Non-cognitive psychopathological symptoms associated with incident mild cognitive impairment and dementia, alzheimer’s type. Neurotox Res. 14, 263–272 (2008). [DOI] [PubMed] [Google Scholar]

[R34] 34.Bickel H & Cooper B Incidence and relative risk of dementia in an urban elderly population: Findings of a prospective field study. Psychol Med. 24, 179–192 (1994). [DOI] [PubMed] [Google Scholar]

[R35] 35.Rolandi E et al. Estimating the potential for dementia prevention through modifiable risk factors elimination in the real-world setting: A population-based study. Alz Res Therapy. 12, 94 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R36] 36.He YL, Zhang XK, & Zhang MY Psychosocial risk factors for Alzheimer’s disease. Hong Kong J Psychiatry. 10, (2000). [Google Scholar]

[R37] 37.Zhou Z et al. The association between loneliness and cognitive impairment among older men and women in China: A nationwide longitudinal study. Int J Environ Res Public Health. 16, 2877 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R38] 38.Wei K et al. Living arrangement modifies the associations of loneliness with adverse health outcomes in older adults: Evidence from the CLHLS. BMC Geriatr. 22, 59 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R39] 39.Rawtaer I et al. Psychosocial risk and protective factors and incident mild cognitive impairment and dementia in community dwelling elderly: Findings from the Singapore Longitudinal Ageing Study. J Alzheimers Dis. 57, 603–611 (2017). [DOI] [PubMed] [Google Scholar]

[R40] 40.Wang H et al. Longitudinal analysis of the impact of loneliness on cognitive function over a 20-year follow-up. Aging Ment Health. 24, 1815–1821 (2020). [DOI] [PubMed] [Google Scholar]

[R41] 41.Terracciano A, Luchetti M, Karakose S, Stephan Y & Sutin AR Loneliness and risk of Parkinson disease. JAMA Neurol. 80, 1138 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R42] 42.Wang F et al. A systematic review and meta-analysis of 90 cohort studies of social isolation, loneliness, and mortality. Nat Hum Behav. 7, 1307–1319 (2023). [DOI] [PubMed] [Google Scholar]

[R43] 43.Yan S et al. Association between sedentary behavior and the risk of dementia: A systematic review and meta-analysis. Transl Psychiatry. 10, 112 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R44] 44.Zhong G, Wang Y, Zhang Y, Guo JJ & Zhao Y Smoking is associated with an increased risk of dementia: A meta-analysis of prospective cohort studies with investigation of potential effect modifiers. PLoS One. 10, e0118333 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R45] 45.Aschwanden D et al. Is personality associated with dementia risk? A meta-analytic investigation. Ageing Res Rev. 67, (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R46] 46.Buecker S, Maes M, Denissen JJA & Luhmann M Loneliness and the Big Five personality traits: A meta–analysis. Eur J Pers. 34, 8–28 (2020). [Google Scholar]

[R47] 47.Perissinotto CM, Stijacic Cenzer I & Covinsky KE Loneliness in older persons: A predictor of functional decline and death. Arch Intern Med. 172, (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R48] 48.Cachón-Alonso L, Hakulinen C, Jokela M, Komulainen K & Elovainio M Loneliness and cognitive function in older adults: Longitudinal analysis in 15 countries. Psychol Aging. 38, 778–789 (2023). [DOI] [PubMed] [Google Scholar]

[R49] 49.Rafnsson SB, Orrell M, d’Orsi E, Hogervorst E & Steptoe A Loneliness, social integration, and incident dementia over 6 years: Prospective findings from the English Longitudinal Study of Ageing. J Gerontol B Psychol Sci Soc Sci. 75, 114–124 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R50] 50.Ferreira RG, Brandão MP & Cardoso MF An update of the profile of older adults with dementia in Europe: Findings from SHARE. Aging Ment Health. 24, 374–381 (2020). [DOI] [PubMed] [Google Scholar]

[R51] 51.Button KS et al. Power failure: Why small sample size undermines the reliability of neuroscience. Nat Rev Neurosci. 14, 365–376 (2013). [DOI] [PubMed] [Google Scholar]

[R52] 52.IntHout J, Ioannidis JPA, Borm GF & Goeman JJ Small studies are more heterogeneous than large ones: A meta-meta-analysis. J Clin Epidemiol. 68, 860–869 (2015). [DOI] [PubMed] [Google Scholar]

[R53] 53.McHugh Power J, Tang J, Kenny RA, Lawlor BA & Kee F Mediating the relationship between loneliness and cognitive function: The role of depressive and anxiety symptoms. Aging Ment Health. 24, 1071–1078 (2020). [DOI] [PubMed] [Google Scholar]

[R54] 54.Cacioppo JT, Hawkley LC & Thisted RA Perceived social isolation makes me sad: 5-year cross-lagged analyses of loneliness and depressive symptomatology in the Chicago Health, Aging, and Social Relations Study. Psychol Aging. 25, 453–463 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R55] 55.O’Rourke HM, Collins L & Sidani S Interventions to address social connectedness and loneliness for older adults: A scoping review. BMC Geriatr. 18, 214 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R56] 56.Joshi P et al. Social connections as determinants of cognitive health and as targets for social interventions in persons with or at risk of Alzheimer’s disease and related disorders: a scoping review. Int Psychogeriatr. 36, 92–118 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R57] 57.Hawkley LC, Thisted RA & Cacioppo JT Loneliness predicts reduced physical activity: Cross-sectional & longitudinal analyses. Health Psychol. 28, 354–363 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R58] 58.Christiansen J, Larsen FB & Lasgaard M Do stress, health behavior, and sleep mediate the association between loneliness and adverse health conditions among older people? Soc Sci & Med. 152, 80–86 (2016). [DOI] [PubMed] [Google Scholar]

[R59] 59.Shiovitz-Ezra S & Parag O Does loneliness ‘get under the skin’? Associations of loneliness with subsequent change in inflammatory and metabolic markers. Aging Ment Health. 23, 1358–1366 (2019). [DOI] [PubMed] [Google Scholar]

[R60] 60.Paul E, Bu F & Fancourt D Loneliness and risk for cardiovascular disease: Mechanisms and future directions. Curr Cardiol Rep. 23, 68 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R61] 61.Maharani A, Pendleton N & Leroi I Hearing Impairment, loneliness, social isolation, and cognitive function: Longitudinal analysis using English Longitudinal Study on Ageing. Am J Geriatr Psychiatry. 27, 1348–1356 (2019). [DOI] [PubMed] [Google Scholar]

[R62] 62.Akhter-Khan SC et al. Associations of loneliness with risk of Alzheimer’s disease dementia in the Framingham Heart Study. Alzheimer’s & Dementia 17, 1619–1627 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R63] 63.Li Y et al. Eight-year trajectories of late-life loneliness and incident dementia: A nationally representative cohort study. Am J Geriatr Psychiatry. 31, 475–486 (2023). [DOI] [PubMed] [Google Scholar]

[R64] 64.Akhter-Khan SC, Prina M, Wong GH-Y, Mayston R & Li L Understanding and addressing older adults’ loneliness: The Social Relationship Expectations Framework. Perspect Psychol Sci. 18, 762–777 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R65] 65.McHugh Power JE et al. Lonely SARTs: Loneliness and sustained attention in the Irish Longitudinal Study of Aging. Neuropsychol Dev Cogn B Aging Neuropsychol Cogn. 27, 197–206 (2020). [DOI] [PubMed] [Google Scholar]

[R66] 66.Yang R et al. Loneliness as a mediator of the impact of social isolation on cognitive functioning of Chinese older adults. Age Ageing. 49, 599–604 (2020). [DOI] [PubMed] [Google Scholar]

[R67] 67.Yu B, Steptoe A, Chen Y & Jia X Social isolation, rather than loneliness, is associated with cognitive decline in older adults: The China Health and Retirement Longitudinal Study. Psychol Med. 51, 2414–2421 (2021). [DOI] [PubMed] [Google Scholar]

[R68] 68.Hughes ME, Waite LJ, Hawkley LC & Cacioppo JT A short scale for measuring loneliness in large surveys: Results from two population-based studies. Res Aging. 26, 655–672 (2004). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R69] 69.Smith J, Ryan LH, Fisher GG, Sonnega A, & Weir DR HRS Psychosocial and Lifestyle Questionnaire 2006–2016. (Survey Research Center, Institute for Social Research, Michigan, 2017). [Google Scholar]

[R70] 70.Radloff LS The CES-D scale: A self-report depression scale for research in the general population. Appl Psychol Meas. 1, 385–401 (1977). [Google Scholar]

[R71] 71.Lewinsohn PM, Seeley JR, Roberts RE & Allen NB Center for Epidemiologic Studies Depression Scale (CES-D) as a screening instrument for depression among community-residing older adults. Psychol Aging. 12, 277–287 (1997). [DOI] [PubMed] [Google Scholar]

[R72] 72.Irwin M, Artin KH & Oxman MN Screening for depression in the older adult: Criterion validity of the 10-item Center for Epidemiological Studies Depression Scale (CES-D). Arch Intern Med. 159, 1701 (1999). [DOI] [PubMed] [Google Scholar]

[R73] 73.Crimmins EM, Kim JK, Langa KM & Weir DR Assessment of cognition using surveys and neuropsychological assessment: The Health and Retirement Study and the Aging, Demographics, and Memory Study. J Gerontol B Psychol Sci Soc Sci. 66B, i162–i171 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R74] 74.Quinn TJ et al. Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE) for the diagnosis of dementia within community dwelling populations. Cochrane Database Syst Rev. CD010079 (2014). [DOI] [PubMed] [Google Scholar]

[R75] 75.Folstein MF, Folstein SE & McHugh PR “Mini-mental state”. A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res. 12, 189–198 (1975). [DOI] [PubMed] [Google Scholar]

[R76] 76.Bassuk SS, Wypij D & Berkmann LF Cognitive impairment and mortality in the community-dwelling elderly. Am J Epidemiol. 151, 676–688 (2000). [DOI] [PubMed] [Google Scholar]

[R77] 77.Creavin ST et al. Mini-Mental State Examination (MMSE) for the detection of Alzheimer’s dementia and other dementias in asymptomatic and previously clinically unevaluated people aged over 65 years in community and primary care populations. Cochrane Database Syst Rev. CD011145 (2016).. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R78] 78.Mejia-Arango S et al. Effect of demographic and health dynamics on cognitive status in Mexico between 2001 and 2015: Evidence from the Mexican Health and Aging Study. Geriatrics. 6, 63 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R79] 79.Lugo-Palacios DG & Gannon B Health care utilisation amongst older adults with sensory and cognitive impairments in Europe. Health Econ Rev. 7, 44 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R80] 80.Giné-Garriga M et al. Is loneliness a predictor of the modern geriatric giants? Analysis from the Survey of Health, Ageing, and Retirement in Europe. Maturitas. 144, 93–101 (2021). [DOI] [PubMed] [Google Scholar]

[R81] 81.Gontié R et al. Relationship between physical activity and incidence of dementia in people aged 50 and over in Europe. Aging Ment Health. 27, 1429–1435 (2023). [DOI] [PubMed] [Google Scholar]

[R82] 82.Aschwanden D, Sutin AR, Luchetti M, Stephan Y & Terracciano A Personality and dementia risk in England and Australia. GeroPsych 33, 197–208 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R83] 83.Hosking DE & Anstey KJ The economics of cognitive impairment: Volunteering and cognitive function in the HILDA Survey. Gerontology. 62, 536–540 (2016). [DOI] [PubMed] [Google Scholar]

[R84] 84.Zhang K & Zhang W Adverse childhood experiences and mild cognitive impairment in later life: Exploring rural/urban and gender differences using CHARLS. J Appl Gerontol. 41, 1454–1464 (2022). [DOI] [PubMed] [Google Scholar]

[R85] 85.Karim J, Weisz R, Bibi Z & ur Rehman S Validation of the eight-item Center for Epidemiologic Studies Depression Scale (CES-D) among older adults. Curr Psychol. 34, 681–692 (2015). [Google Scholar]

[R86] 86.Kim H, Park S-M, Jang S-N & Kwon S Depressive symptoms, chronic medical illness, and health care utilization: Findings from the Korean Longitudinal Study of Ageing (KLoSA). Int Psychogeriatr. 23, 1285–1293 (2011). [DOI] [PubMed] [Google Scholar]

[R87] 87.Ware JE & Gandek B Overview of the SF-36 Health Survey and the International Quality of Life Assessment (IQOLA) Project. J Clin Epidemiol. 51, 903–912 (1998). [DOI] [PubMed] [Google Scholar]

[R88] 88.Castro-Costa E et al. Ascertaining late-life depressive symptoms in Europe: An evaluation of the survey version of the EURO-D scale in 10 nations. The SHARE project. Int J Methods Psychiatr Res. 17, 12–29 (2008). [DOI] [PMC free article] [PubMed] [Google Scholar]

[R89] 89.Tabachnick BG & Fidell LS Using multivariate statistics. (Tabachnick, Tabachnick, 2013). [Google Scholar]

[R90] 90.Viechtbauer W Conducting meta-analyses in R with the metafor Package. J Stat Soft. 36, (2010). [Google Scholar]

[R91] 91.Stanley TD & Doucouliagos H Meta-regression approximations to reduce publication selection bias. Res Synth Methods. 5, 60–78 (2014). [DOI] [PubMed] [Google Scholar]

[R92] 92.Simonsohn U, Nelson LD & Simmons JP P-curve: A key to the file-drawer. J Exp Psychol Gen. 143, 534–547 (2014). [DOI] [PubMed] [Google Scholar]

[R93] 93.Simonsohn U, Simmons JP & Nelson LD Better P-curves: Making P-curve analysis more robust to errors, fraud, and ambitious P-hacking, a reply to Ulrich and Miller (2015). J Exp Psychol Gen. 144, 1146–1152 (2015). [DOI] [PubMed] [Google Scholar]

PERMALINK

A Meta-analysis of Loneliness and Risk of Dementia using Longitudinal Data from >600,000 Individuals

Martina Luchetti, Ph.D.

Damaris Aschwanden, Ph.D.

Amanda A Sesker, Ph.D.

Xianghe Zhu, Ph.D.

Páraic S O’Súilleabháin, Ph.D.

Yannick Stephan, Ph.D.

Antonio Terracciano, Ph.D.

Angelina R Sutin, Ph.D.

Abstract

Introduction

Results

Coordinated Analysis of Aging Cohort Studies

Table 1.

Meta-analysis

Fig. 1: Flow diagram for identifying eligible studies.

Primary outcome = Dementia

Fig. 2: Forest plot for all-cause dementia.

Sensitivity Analysis.

Publication Bias.

Meta-regressions.

Table 2.

Cause-specific dementia.

Fig. 3: Forest plot for Alzheimer’s and vascular dementia.

Secondary outcome = Cognitive Impairment

Fig. 4: Forest plot for cognitive impairment.

Sensitivity analysis.

Publication Bias.

Meta-Regressions.

Discussion

Strengths, Limitations, and Future Directions

Conclusions

Methods

Aging Cohort Studies

Loneliness.

Cognitive Status.

Control variables.

Statistical Analysis.

Meta-analysis

Literature Search Strategy.

Selection Criteria and Quality Assessment.

Statistical Analysis.

Extended Data

Extended Data Fig. 1. Forest plot of fully adjusted models.

Extended Data Fig. 2. Funnel plots to evaluate publication bias.

Extended Data Table. 1.

Extended Data Table. 2.

Extended Data Table. 3.

Extended Data Table. 4.

Extended Data Table. 5.

Extended Data Table. 6.

Supplementary Material

Acknowledgments

Footnotes

Contributor Information

Data availability

References

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases