How loneliness and social isolation are linked to cognitive decline among older adults? A systematic review of underlying physiological and psychobehavioral mechanisms

Abstract

Despite empirical evidence linking loneliness and social isolation to cognitive decline (CD) in later life, the underlying physiological and psychobehavioral mechanisms remain unclear. We conducted a systematic review to synthesize relevant evidence. Studies examining psychobehavioral pathways required mediation analyses, whereas studies on physiological pathways were included when mediation was examined or when outcomes are dementia‐related neuropathological changes. Thirty‐eight articles were included. Evidence pointed to multiple physiological mechanisms, including Alzheimer's disease pathology, brain structural differences, inflammation, cardiovascular risks, neurotrophic factors, hypothalamic–pituitary–adrenal (HPA) ‐axis function, gene expression, and general health, although findings were mixed. Psychobehavioral pathways, including depressive symptoms, neuroticism, lower sense of control, sleep disturbance, and lower engagement in social or leisure activities, mediated the association. This review identified mechanistic targets that may be leveraged in non‐pharmacological interventions. Future intervention programs aimed at reducing social isolation and loneliness to delay CD could investigate whether they impact physiological and psychobehavioral pathways summarized here.

Highlights

• Physiological pathways include Alzheimer's disease (AD) pathology, inflammation, and hypothalamic–pituitary–adrenal (HPA) ‐axis function.

• Psychobehavioral mediators include depression, sleep, and physical inactivity.

• Proposed integrated biopsychosocial framework for future mechanistic research.

• Additional research is needed to test mediation for physiological pathways.

• Identified mechanistic targets for future non‐pharmacological intervention programs.

1. BACKGROUND

Health and social relationships are critical components that fundamentally influence how individuals age, and those two domains are interrelated. ¹ , ² For instance, aspects of social disconnection (e.g., social isolation and loneliness) have been increasingly recognized as modifiable risk factors for the development of Alzheimer's disease and related dementias (ADRD), ³ which offers an opportunity for disease prevention given the prevalent experience of social disconnectedness in later life. About one‐third of community‐dwelling older adults report frequent loneliness, ³ and the estimated global prevalence of social isolation in this population is 25%. ⁴ A meta‐analysis with longitudinal data from 600,000+ older adults found loneliness is associated with a 31% higher risk of all‐cause dementia, independent of social isolation or depression. ⁵ Social connection, the opposite of social isolation, has been linked to reduced risk of mild cognitive impairment (MCI) and dementia. ⁶ , ⁷ Further, according to the 2024 Lancet Commission on Dementia Prevention report, about 5% of dementia cases could be attributed to social isolation and loneliness, which corresponds to about 2.75 million patients globally and 350,000 in the U.S. alone. ⁸ Despite extensive empirical evidence suggesting both social isolation and loneliness are separate risk factors for cognitive decline (CD) and possible subsequent dementia onset, the underlying mechanisms are less clear. ⁷ , ⁹ The current systematic review is set to address this gap by synthesizing evidence on the physiological and behavioral pathways linking loneliness and social isolation to CD.

Because social isolation and loneliness are related but distinctly different concepts, it is important to examine them together when considering pathways to CD. Social isolation refers to having few or no social relationships or interactions, ⁴ whereas loneliness refers to the subjective feeling that one's social connections are insufficient in quality or quantity. ¹⁰ In other words, social isolation captures a more objective lack of social contact, while loneliness represents the subjective experience of social disconnection. Social isolation and loneliness are positively associated with each other, but one can be isolated but not lonely (i.e., solitude) and vice versa (i.e., lonely in crowd). Further, social isolation and loneliness have been found to have an independent relationship with CD. ¹¹ , ¹² Synthesizing evidence related to both constructs in one review can contribute to identifying shared and different underlying pathways to CD, and informing mechanisms targeted intervention for older adults experiencing social disconnection and at higher risk of CD.

RESEARCH IN CONTEXT

1. Systematic review: The authors reviewed literature using traditional sources (e.g., PubMed). Loneliness and social isolation were associated with cognitive decline (CD) in later life. However, the physiological and psychobehavioral mechanisms underlying these associations remain unclear, and no prior systematic review has synthesized the relevant evidence.

2. Interpretation: Findings indicate that loneliness and social isolation are linked to CD through physiological changes (e.g., Alzheimer's pathology, brain structural alterations, inflammation, cardiovascular risks, neurotrophic factors, HPA‐axis function, and general health), and psychobehavioral pathways (e.g., depressive symptoms, neuroticism, low sense of control, sleep disturbance, and reduced engagement in social or leisure activities).

3. Future directions: Future studies using time‐sensitive measures and longitudinal study design may clarify the temporal dynamics of mediating pathways. Intervention programs aimed at reducing social isolation and loneliness could investigate whether they impact pathways summarized here. A combined research approach on how psychobehavioral may interact with physiological vulnerabilities can reveal more contextualized risks for CD.

A biobehavioral framework is useful for understanding how social disconnection relates to cognitive outcomes, since it recognizes that cognition is shaped by biological, psychological, and behavioral processes. Biological pathways emphasize the impact that loneliness and social isolation can have on the physical body, which can increase risk for CD. Potential biological mechanisms include factors related to stress physiology, including dysregulation of the hypothalamic–pituitary–adrenal (HPA) axis. For example, loneliness is related to heightened cortisol, ¹³ which may contribute to CD due to the high concentration of glucocorticoid receptors in the hippocampus and prefrontal cortex–brain regions critical for cognitive function. ¹⁴ Additionally, loneliness is associated with altered neural networks, which could accelerate neurodegenerative processes over time and contribute to CD. ¹⁵ Living alone, a proxy of social isolation, is related to a flattened diurnal cortisol pattern and higher C‐reactive protein (CRP) levels in adults. ¹⁶ Psychobehavioral pathways emphasize the negative impact that loneliness and social isolation can have on socioemotional wellbeing and behavior, which can increase risk for CD. Potential psychobehavioral pathways include poor mental health and maladaptive behaviors. For example, depressive symptoms and suboptimal sleep and physical activity patterns may all contribute to diminished cognitive reserve and reduced resilience to cognitive aging. ¹⁷ , ¹⁸ , ¹⁹ , ²⁰ Framing this review within a biobehavioral perspective highlights the multiple pathways that could connect social disconnection to CD and sets the stage for future work examining their interplay.

Although previous reviews established the associations between social isolation, loneliness, and CD, ⁵ , ²¹ , ²² , ²³ , ²⁴ this systematic review aims to address a critical gap in understanding mechanisms through which social isolation and loneliness can be linked to CD among older adults. While previous evidence supports social isolation and loneliness as risk factors that precede and contribute to CD, ⁶ , ⁷ we acknowledge an opposing view that suggests that early cognitive changes manifest as social withdrawal, thereby increasing loneliness and social isolation, with loneliness being a prodromal symptoms of dementia. ²⁵ , ²⁶ By synthesizing mechanistic evidence, particularly from longitudinal studies incorporating ADRD biomarkers (e.g., amyloid‐β, tau, and brain atrophy), the current review evaluated empirical evidence that supports or opposes the plausibility of proposed pathways. Research areas with shared interest in this topic—such as Neurology, Psychology, Gerontology, and Public Health—have tended to operate in silos, limiting opportunities for cross‐disciplinary insights. This review integrates findings across physiological and psychobehavioral pathways, producing a more comprehensive understanding of the mechanisms and promoting future collaborative research agendas. This review project was guided by the following review questions.

1. What are the underlying physiological and psychobehavioral pathways that can explain the association between loneliness and CD among older adults?

2. What are the underlying physiological and psychobehavioral pathways that can explain the association between social isolation and CD among older adults?

2. METHODS

In the current review, we used the terms mechanisms and pathways interchangeably. CD was defined as a lower cognitive test score or an MCI/dementia incident. The systematic review protocol has been registered at the International Prospective Register of Systematic Reviews (PROSPERO; registration number: CRD42023404376). Throughout the review process, we amended the protocol three times to make changes to core team members and update review progress. This review was conducted following the Preferred Reporting Items for Systematic Reviews and Meta‐Analysis (PRISMA) guidelines [ ²⁷ ]. The completed PRISMA checklist for this review can be found in Table S1. We used Covidence for reference management, screening, data extraction, and risk of bias assessment.

For research examining psychobehavioral pathways, studies that investigated mediating pathways of the associations between social isolation, loneliness, and CD were included. To identify a broader range of potential mechanisms, studies that examined physiological pathways were not required to conduct a mediation analysis as long as the physiological outcomes directly addressed dementia pathology, for example, amyloid‐beta, tau, neuroinflammation, change in brain volume, white matter hyperintensity (WMH). We chose to include studies investigating brain physiological changes associated with social isolation and loneliness, recognizing that this is a rapidly developing research area where mediation analyses with cognitive outcomes are often lacking. The advantages and limitations of this decision are outlined in Section 4.6.

2.1. Search strategy

A systematic search of five databases, including PubMed, Web of Sciences, EMBASE, PsycINFO, and MEDLINE were conducted from inception to October 2024. We manually searched the reference list of all included studies to identify additional studies. The following search terms reflecting key concepts of the review were connected with AND logic. The same set of search terms was applied across databases.

1. Loneliness and social isolation: Lonel* OR “social isolation” OR “subjective isolation” OR “social withdrawal” OR “feel isolated” OR “felt isolated” OR “feeling* of isolation” OR aloneness OR “felt alone” OR “feel* alone” OR “social alienation” OR “social exclusion”

2. Aging: aged OR elderly OR “older adult” OR aging OR geriatric OR ageing OR old

3. Cognition: Cogniti* OR memory OR executive OR speed OR visuospatial OR semantic OR attention OR psychomotor OR Mild Cognitive Impairment OR MCI OR Alzheimer's disease OR Dementia OR Neuropsychology assessment OR UDS

4. Mechanisms: Mechanisms OR underpin* OR pathway* OR Mediati* OR Moderati* OR Amyloid OR Tau OR Neuro inflammation OR cerebrospinal fluid biomarker OR CSF biomarker OR fluid biomarker OR fMRI OR structure OR TDP‐43 OR Lewy Bodies OR Brain‐derived neurotrophic factor OR BDNF OR Neurofilament Light chain OR NfL OR white matter hyperintensities OR Head CT OR MRI OR Plasma biomarker* OR Blood biomarker* OR positron emission tomography OR PET OR Pathology OR Stroke OR Vascular dementia OR Vascular cognitive impairment OR Atrophy OR Hippocampus OR Braak OR CERAD OR Neurofibrillary tangles

2.2. Selection criteria

Studies were included if they met the following criteria: (1) included middle‐ and older‐aged individuals (50+ years) from the general community‐based population; (2) examined loneliness or social isolation as the exposure variable; (3) examined CD as defined by a lower cognitive test scores and MCI/dementia incidence; (4) longitudinal and cross‐sectional mediation analysis for psychobehavioral pathways; (5) longitudinal and cross‐sectional studies examining the association between social isolation, loneliness, and dementia‐related neuropathological changes (including but not limited to amyloid, tau, neuroinflammation, and TAR DNA‐binding protein 43 (TDP‐43), see the full list of neuropathological changes included in search term 4. Mechanisms). Studies that examined neuropathology did not have to include CD as an outcome to be included; (6) peer reviewed articles published in English.

Exclusion criteria were: (1) qualitative studies; (2) study conducted with specific clinical populations, such as older adults with severe depression.

2.3. Data collection and analysis

Four (K.Y., J.K., K.V.B., and H.T.) reviewers independently screened and selected studies for inclusion. At least two reviewers screened each article. We used Covidence to coordinate and record decisions. The search results were screened in two stages: title and abstract screening and full‐text screening. Disagreements were discussed in biweekly team meetings. A third reviewer was invited to reconcile different opinions between the primary reviewers when necessary.

For all included studies, four reviewers (K.Y., J.K., K.V.B., and H.T.) independently extracted publication information (author, year), setting, countries where the research was conducted, study design, sample size, participant demographics, recruitment methods, measurement instruments used for loneliness, social isolation, and cognition. For included longitudinal studies, we also extracted information regarding the follow‐up period, number of assessments, and attrition. If the included study examined ADRD biomarkers, we extracted the method used for assessing biomarkers (e.g., blood, positron emission tomography [PET] imaging, and cerebrospinal fluid [CSF]) and whether the biomarkers assessed were specific to various brain regions. Data extraction for each article was completed by two independent reviewers. KY cross‐checked all extracted data and any disagreements were resolved in regular team meetings. Each article reached consensus for data extraction.

2.4. Study quality and risk of bias

We used the Joanna Briggs Institute (JBI) risk of bias checklist for cross‐sectional studies and checklist for cohort studies to evaluate the quality of evidence for included articles. ²⁸ Following PRISMA recommendations, we did not compute an overall quality rating for studies. Instead, we reviewed quality in separate areas, such as sample selection, study setting, variable measurement, confounding adjustment, and statistical analysis. Four reviewers (K.Y., J.K., K.V.B., and H.T.) independently assessed the quality of each included study. Each included study's risk of bias was assessed by two reviewers. Any disagreements were reconciled in regular team meetings.

3. RESULTS

3.1. Study selection

We identified 4314 unique articles, 4300 from the systematic search, and 14 from searching the reference lists. After title and abstract screening, the team assessed 104 full‐text articles for eligibility. Finally, 37 unique articles met the inclusion criteria. The process of study selection protocol and reasons for exclusion are summarized in the PRISMA flow diagram (Figure 1).

FIGURE 1.

Preferred Reporting Items for Systematic Reviews and Meta‐Analysis (PRISMA) flow diagram for study selection.

3.2. Study and participant characteristics

Tables 1, 2, 3, 4 summarize the study characteristics of each included study. For underlying pathways between loneliness and CD, the current review included 19 studies on physiological pathways (Table 1) and seven studies on psychobehavioral pathways (Table 2). Regarding social isolation and CD, eight studies examined physiological pathways (Table 3) and seven investigated psychobehavioral pathways (Table 4). Detailed measurement information and key findings of each study can be found in Tables S2–S5. Kim et al. ²⁹ and Dabiri et al. ³⁰ examined both physiological and psychobehavioral pathways for loneliness (listed in Tables 1and 2), and Duffner et al. ³¹ reported psychobehavioral pathways of both social isolation and loneliness (Tables 2 and 4).

TABLE 1.

Characteristics of included studies examined physiological pathways of the proposed association between loneliness and cognitive decline (N = 18).

Author (year)	Location	Sample size	Sociodemographic	Follow‐up	Measurement of loneliness	Cognitive domains	Physiological pathways
Cross‐sectional studies (N = 8)
Imai et al. (2022) 65	Japan	110	Sex: 69.1% female; Age:75.7 ± 7.3; Education: NR; Race/Ethnicity: Japanese	NA	R‐UCLA‐LS	NA	Regional GMV
Zhang et al. (2022) 53	Japan	176	Sex: 62.5% female; Age: 77.8 ± 9.1; Education: 12.8 ± 2.5 years; Race/Ethnicity: Japanese	NA	Single yes/no question: “Do you often feel lonely?”	NA	Regional GMV
Düzel et al. (2019) 66	Germany	319	Sex: 38.3% female; Age: 70.0 ± 3.7; Education: 14.2 ± 2.78; Race/Ethnicity: NR	NA	7‐Item UCLA‐LS	NA	Regional GMV
Uquillas (2018) 38	USA	117	Sex: 59.0% female; Age: 76.0 ± 6.2; Education: NR; Race/Ethnicity: NR	NA	3‐Item UCLA‐LS	Logical memory	Tau pathology
Chen et al. (2022) 68	USA	45	Sex: 64.34% female; Age: 63.4 ± 6.6; Education: NR; Race/Ethnicity: NR	NA	20‐Item UCLA‐LS	NA	Cerebral blood flow
Montoliu et al. (2019) 71	Spain	86	Sex: 52.3% female; Age: 67.4 ± 4.4; Education: 58.8% had some college education; Race/Ethnicity: NR	NA	Spanish R‐UCLA‐LS	Global cognition, processing speed, executive function, working memory, verbal memory	Salivary cortisol
Lao et al. (2024) 35	USA	680	Sex: 68.0% female; Age: 88.8 ± 6.5; Education: 14.8 ± 2.9; Race/Ethnicity: 94% White, 6% Black	NA	5‐Item modified DJG‐LS	NA	Microinfarcts
Park et al. (2023) 36	Korea	173	Sex: 59.9% female; Age: 71.9 ± 6.8; Education: 7.2 ± 4.3; Race/Ethnicity: Korean	NA	Korean R‐UCLA‐LS	Episodic memory	WMH, GMV, bilateral hippocampal volumes
Longitudinal studies (N = 10)
Van der Velpen et al. (2024) 52	Netherlands	4099	Sex: 57.0% female; Age: 71.5 ± 7.3; Education:14.2% higher vocational education or university; Race/Ethnicity: NR	Median 6.5 years	Single yes/no question: feeling lonely > 1 day/week	Global cognition	Plasma Aβ40, Aβ42, total Tau, NfL
Kim et al. (2020) 29	USA	3005	Sex: 51.6% female; Age: 69.3 ± 7.9; Education: NR; Race/Ethnicity: 76.37% Non‐Hispanic White, 16.94% non‐Hispanic Black, 1.2% Asian, 4.76% other	10 years	3‐Item UCLA‐LS	Global cognition	Multisystemic physiological risk
Canli et al. (2017) 33	USA	26	Sex: 46.2% female; Age: 86.9 ± 6.7; Education: 14.7 ± 3.3; Race/Ethnicity: Non‐Hispanic White	2.3 years	5‐Item modified DJG‐LS	NA	Differentially expressed genes
Canli et al. (2018) 32	USA	181	Sex: 65.8% female; Age: 89.5 ± 6.2; Education:14.7 ± 2.7; Race/Ethnicity: NR	5 years	5‐Item modified DJG‐LS	NA	RNA expression
Zhong et al. (2017) 50	China	14,199	Sex: 55.7% female; Age: 85.0 ± 11.3; Education: NR; Race/Ethnicity: NR	10 years	Single question ask how often one felt lonely	Global cognition	Number of chronic conditions
Yu and Ng (2023) 69	USA	7037	Sex: 57.4% female; Age: 68.0(SD NR); Education: 44.97% had some college education; Race/Ethnicity: 73.7% non‐Hispanic White	8 years	3‐Item UCLA‐LS	Global cognition	HbA1c, HDL and LDL cholesterol, CRP, and cystatin C
Wilson et al. (2007) 34	USA	90	Sex: NR; Age: 86.1 ± 5.8; Education: NR; Race/Ethnicity: NR	Up to 4 years	5‐Item modified DJG‐LS	Clinical diagnosis Global cognition	Aβ, Tau, cerebral infarction
Duan et al. (2017) 67	China	219	Sex: 50.7% female; Age: 69.9 ± 5.5; Education: 9.0 ± 4.3; Race/Ethnicity: NR	5.2 years	20‐Item Chinese UCLA‐LS	Global cognition	Periventricular WMH, deep WMH, and WMH
Dabiri et al. (2024) 30	USA	2130	Sex: 69.6% female; Age: 80.1 ± 7.6; Education: 14.9 ± 3.34; Race/Ethnicity: NR	up to 24 years	5‐Item modified DJG‐LS	Global cognition	Total GMV, BDNF, Pro‐inflammatory cytokines
Sole‐Padulles et al. (2022) 42	Sweden, Germany	1468	Sex: Swedish cohort 57.3% female, German cohort 50.6% female; Age: Swedish cohort 64.0 ± 3.9, German cohort 70.6 ± 8.1; Education: Swedish cohort 10.1 ± 3.9, German cohort 14.1 ± 2.9 years; Race/Ethnicity: NR	Swedish cohort 4 years German cohort 2.3 ± 0.5 years	Harmonized measures by classifying the loneliness scores to high, medium, and low categories	Verbal episodic memory	Hippocampal volume and cortical thickness

Abbreviations: Aβ, amyloid beta; BDNF, brain‐derived neurotrophic factor; CRP, C‐reactive protein; DJG‐LS, DeJong Gierveld Loneliness Scale; GMV, gray matter volume; HbA1c, hemoglobin A1c; HDL, high‐density lipoprotein; LDL, low‐density lipoprotein; NA, not applicable; NfL, neurofilament light; NR, not reported; R‐UCLA‐LS, reviewed UCLA Loneliness Scale; UCLA, UCLA Loneliness Scale; USA, United States of America; WMH, white matter hyperintensity.

TABLE 2.

Characteristics of included studies examined psychobehavioral pathways of the association between loneliness and cognitive decline (n = 7).

Author (year)	Location	Sample size	Sociodemographic	Follow‐up	Measurement of loneliness	Cognitive domains	Psychobehavioral pathways
Cross‐sectional analysis (n = 3)
Foong et al. (2018) 72	Malaysia	2322	Sex: 52.0% female; Age: 69.1 ± 6.2; Education: 5.1 ± 4.0; Race/Ethnicity: NR	NA	3‐Item UCLA‐LS	Global cognition	Neuroticism
Du et al. (2024) 74	China	4772	Sex: 50.7% female; Age: 69.1 ± 6.2; Education: 11.0 ± 3.1; Race/Ethnicity: Chinese	NA	R‐UCLA‐LS	Global cognition, memory, language, attention, execution, visual space	Leisure activity
Yang et al. (2021) 73	Singapore	69	Sex: 76.8% female; Age: 70.4 ± 7.3; Education: on a scale from 1 (no school) to 12 (doctoral degree), the sample mean was 3.1; Race/Ethnicity: NR	NA	3‐Item UCLA‐LS	Global cognition Working memory	Sense of control, perceived constraint, personal mastery
Longitudinal studies (n = 4)
Dabiri et al. (2024) 30	USA	2130	Sex: 69.6% female; Age: 80.1 ± 7.6; Education: 14.9 ± 3.34; Race/Ethnicity: NR	Up to 24 years	5‐Item modified DJG‐LS	Global cognition	Depressive symptoms
Duffner et al. (2024) 31	England	7536	Sex: 55.5% female; Age: 66.3 ± 9.7; Education:34.3% no formal education, 21.9% secondary education, 37.0% college or higher education; Race/Ethnicity: NR	11.8 years	4‐Items from UCLA‐LS‐V3 & “felt lonely much of the time during the past week (Y/N)”	Clinical diagnosis	Depressive symptoms
Kim et al. (2020) 29	USA	3005	Sex: 51.6% female; Age: 69.3 ± 7.9; Education: NR; Race/ Ethnicity: 76.37% non‐Hispanic White, 16.94% non‐Hispanic Black, 1.20% Asian	10 years	3‐Item UCLA‐LS	Global cognition	Depressive symptoms, social participation
Power et al. (2020) 43	Ireland	7433	Sex: 53.4% female; Age: 64.0 (SD NR); Education:30.7% no formal education, 40.2% secondary education, 29.1% college or higher education; Race/Ethnicity: Irish	2–4 years	5‐Item UCLA‐LS	Global cognition	Depressive symptoms, anxiety

Abbreviations: DJG‐LS, DeJong Gierveld Loneliness Scale; NA, not applicable; NR, not reported; R‐UCLA‐LS, reviewed UCLA Loneliness Scale; UCLA, UCLA Loneliness Scale; UCLA‐LS‐V3, UCLA Loneliness Scale version 3; Y/N, yes/no.

TABLE 3.

Characteristics of included studies examined physiological pathways of the proposed association between social isolation and cognitive decline (n = 8).

Author (year)	Location	Sample size	Sociodemographic	Follow‐up	Measurement of social isolation	Cognitive domains	Physiological pathways
Cross‐sectional studies (n = 4)
Qi et al. (2023) 55	USA	2535	Sex: 50.0% female; Age:71.1 ± 7.8; Education: 40.1% less than high school 24% high school; 35.9% more than high school; Race/Ethnicity:59.6% non‐Hispanic White, 14.8% non‐Hispanic Black, 23.4% Hispanic, 2.2% other	NA	Berkman–Syme Social Network Index	Processing speed, attention, working memory	CRP, fibrinogen, albumin
Hirabayashi et al. (2023) 59	Japan	8896	Sex: 57.0% female; Age: 73.0 ± 6.0; Education: NR; Race/Ethnicity: Japanese	NA	Single question: “How often are you in contact with relatives and friends who do not live with you”?	NA	Total and regional brain volumes, intracranial volume, white matter lesion volume
Chang et al. (2024) 58	USA	69	Sex: 65.2% female; Age: 69.0 ± 7.1; Education: 16.7 ± 2.2 years; Race/Ethnicity: NR	NA	Cohen's Social Network Index	NA	Cerebrovascular reactivity
Salinas et al. (2017) 56	USA	3294	Sex: 56.0% female; Age:65(SD NR); Education: 7% less than high school, 32% high school, 28% some college, 33% college graduate; Race/Ethnicity: NR	NA	Berkman–Syme Social Network Index	Clinical diagnosis	BDNF
Longitudinal studies (n = 4)
Lammer et al. (2023) 48	Germany	1992	Sex: 46.0% female; Age: 67.0 ± 7.0; Education: 13% had a tertiary degree; Race/Ethnicity: NR	6–7 years	LSNS‐6	Global cognition	Hippocampal volume, cortical thickness
Biddle et al. (2019) 39	USA	217	Sex: 60.8% female; Age: 73.5 ± 6.1; Education: 16.01 ± 2.92; Race/Ethnicity: NR	3 years	Time spent per week on social activities	Episodic memory, executive function, global cognition	Aβ deposition
Khalaila et al. (2024) 45	Europe	55,340	Sex: 55.8% female; Age: 65.3 ± 9.8; Education: 75.5% had secondary education or below; Race/Ethnicity: NR	4 years	Steptoe Social Isolation Index	Global cognition	Number of missing teeth
Murayama et al. (2025) 44	Japan	279	Sex: 52.4% female; Age: 73.0 ± 5.1; Education: 35.3% ≤9 years; Race/Ethnicity: Japanese	4 years	Social contact less than once per week	NA	Hippocampal volume, total GMV

Abbreviations: Aβ, amyloid beta; BDNF, brain‐derived neurotrophic factor; CRP, C‐reactive protein; GMV, gray matter volume; LSNS‐6, 6‐item Lubben Social Network Scale; NA, not applicable; NR, not reported; SD, standard deviation; USA, United States of America.

TABLE 4.

Characteristics of included studies examined psychobehavioral pathways of the association between social isolation and cognitive decline (n = 7).

Author (year)	Location	Sample size	Education	Follow‐up	Measurement of social isolation	Cognitive domains	Psychobehavioral pathways
Cross‐sectional (n = 3)
Jang et al. (2021) 75	USA	2061	Sex: 66.8% female; Age: 73.2 (SD NR); Education: greater than a high school education: 39.7%; Race/Ethnicity: Korean	NA	Korean LSNS‐6	Objective and subjective cognitive impairment	Loneliness
Yang et al. (2020) 37	China	7410	Sex: 49.7% female; Age: 68.4 ± 7.0; Education: 93.0% less than secondary; 5.1% secondary and vocational training; 2.0% Tertiary education; Race/Ethnicity: Chinese	NA	Living alone, social activity, weekly adult children contact, caregiving for grandchildren	Global cognition	Loneliness
Bai et al. (2024) 54	China	496	Sex: 52.2% female; Age: 70.1 ± 6.9; Education: 24.4% college or above; Race/Ethnicity: Chinese	NA	Chinese LSNS‐6	Global cognition	Depressive symptoms
Longitudinal studies (n = 4)
Duffner et al. (2024) 31	England	7536	Sex:55.5% female; Age:66.3 ± 9.7; Education:37.0% had college education; Race/Ethnicity: NR	15 years	Contact with children, other family, and friends, social engagement, negative social experiences, loneliness	Clinical diagnosis	Depressive symptoms
Qi et al. (2023) 57	USA	5753	Sex:57.8% female; Age:65.1 ± 9.5; Education:13.9% less than college, 34.8% some college, 51.3% college or above; Race/Ethnicity: 81.4% non‐Hispanic White,10.3% non‐Hispanic Black, 6.7% Hispanic, 1.6% Others	8 years	Steptoe Social Isolation Index	Global cognition	Sleep disturbance
Zhou et al. (2020) 41	China	6998	Sex: 51.2% female; Age: 81.0 ± 10.1; Education: 2.62 ± 3.68 years; Race/Ethnicity: Chinese	3 years	Marital status, living alone, availability of help and confidant, social participation	Global cognition	Psychological wellbeing
Hopper et al. (2024) 46	Canada	21,491	Sex: 50.0% female; Age: 73 (SD NR); Education: 7 years (SD NR); Race/Ethnicity: 97% White	3 years	The Canadian Longitudinal Study on Aging Social Isolation Index	Executive function, processing speed, memory	Physical activity

Abbreviations: LSNS‐6, 6‐item Lubben Social Network Scale; NA, not applicable; NR, not reported; SD, standard deviation.

Two‐thirds of the studies included in this review were published within the past 5 years, reflecting the rapid growth of the research topic. The studies were conducted in a diverse array of countries spanning North America, Asia, and Europe. Among the 30 studies reporting education, approximately half included samples with relatively higher educational attainment, reflected in mean education levels of 14–16 years ³⁰ , ³² , ³³ or substantial proportions with college‐level education. ³⁴ , ³⁵ Geographic differences in educational attainment were observed among these studies. Studies conducted in Asian populations (n = 10) tended to report lower education (e.g., <10–12 years or majority ≤ secondary), ³⁶ , ³⁷ whereas studies conducted in North America (n = 12) and Europe (n = 8) more often reported higher average education (14–16 years) or substantial proportions with college‐level education, with some variability. ³⁸ Eighteen out of the 37 included studies utilized data from larger aging cohort studies using population‐based systematic sampling. Some studies were conducted as part of a longitudinal observational study that recruited participants from a specific geographic region (e.g., Rush Memory Aging Project and Harvard Aging Brian Study). ³⁰ , ³² , ³³ , ³⁴ , ³⁵ , ³⁸ , ³⁹ , ⁴⁰ Sample sizes for the included studies ranged from 26 to 55,340. The follow‐up period among longitudinal studies ranged from 2 to 24 years, with nine studies with follow‐up of less than 5 years, ³³ , ³⁴ , ³⁹ , ⁴¹ , ⁴² , ⁴³ , ⁴⁴ , ⁴⁵ , ⁴⁶ and eight studies of more than 10‐year of follow‐up. ³⁰ , ³¹ , ⁴⁷ , ⁴⁸ , ⁴⁹ , ⁵⁰ , ⁵¹ , ⁵²

The most commonly used loneliness measure was the University of California Los Angeles Loneliness Scale (UCLA‐LS, n = 16), followed by the modified DeJong Gierveld Loneliness Scale (DJG‐LS, n = 5). Studies varied on which items they used from the full UCLA‐LS, ranging from 3 to 20 items. Three studies used a single binary item to assess loneliness. ⁵⁰ , ⁵² , ⁵³ Of the 16 studies that examined social isolation, three measured social isolation with the 6‐item Lubben Social Network Scale (LSNS‐6), ⁴¹ , ⁴⁸ , ⁵⁴ two with Berkman–Syme Social Network Index, ⁵⁵ , ⁵⁶ two with the Steptoe Social Isolation Index, ⁴⁵ , ⁵⁷ and one with Cohen's social network index. ⁵⁸ One study measured social isolation using a single‐item indicating frequency of contact with family and friends who do not live with them. ⁵⁹ The remaining studies did not explicitly state a type of social network index and operationalized social isolation using time spent per week on social activities, ³⁹ or a single or multiple latent variables constructed with various factors, such as frequency of contact with children, friends, and other family members, and frequency of social participation. ³¹ , ³⁷

Cognitive function was assessed using a wide range of approaches across studies. More thorough information on measurements can be found in Tables S2–S5. Briefly, global cognitive function was most commonly measured using the Mini‐Mental State Examination (MMSE, n = 8), ⁶⁰ the Montreal Cognitive Assessment (MoCA, n = 4), ⁶¹ and the Telephone Interview for Cognitive Status (TICS, n = 3). ⁶² Six studies measured global cognition with a composite score from a number of domain specific cognition measures. ³⁰ , ³⁴ , ⁴³ , ⁴⁵ , ⁴⁸ , ⁵² Common ,  ,  ,   domain‐specific cognitive functions assessed across studies included attention, working memory, visuospatial ability, executive function, processing speed, language (e.g., verbal fluency), and memory (e.g., episodic, semantic). Cognitive status (normal, MCI, or dementia) was categorized according to Clinical Dementia Rating (CDR), ⁶³ Diagnostic and Statistical Manual of Mental Disorders (DSM) ‐IV criteria, ⁶⁴ one study (Duffner et al.) used self‐report of a physician diagnosis combined with the Informant Questionnaire on Cognitive Decline in the Elderly. ³¹

The physiological and psychobehavioral underlying pathways for the associations between loneliness, social isolation, and CD identified in this systematic review are summarized in Figure 2. Below we provide a summary of relevant evidence for the two predictors, that is, loneliness and social isolation.

FIGURE 2.

Physiological and psychobehavioral pathways that separately explain the associations between loneliness, social isolation, and cognitive decline. NS, not statistically significant relationship; ↑, positive association; ↓, negative association. Aβ, amyloid beta‐protein; BDNF, brain‐derived neurotrophic factor; CRP, C‐reactive protein; HbA1C, hemoglobin A1C; HPA‐axis, hypothalamic–pituitary–adrenal axis; IL‐6, interleukin‐6; IL‐1b, interleukin‐1 beta; TNF‐α, tumor necrosis factor alpha.

3.3. Loneliness and cognitive decline

3.3.1. Physiological pathways

The current review yielded several physiological pathways linking loneliness to CD, including (1) Alzheimer's disease (AD) pathology, (2) brain structural changes, (3) cardiovascular risks, (4) neurotrophic factors, (5) AD‐related gene expression, (6) general health indicators, (7) inflammation, and (8) HPA‐axis functioning, with both significant and null findings reported.

1. AD pathology (cross‐sectionally related to loneliness, not related to loneliness according to longitudinal evidence): Longitudinal evidence does not support a direct association between loneliness and AD neuropathological markers, ³⁴ , ⁵² and while one cross‐sectional study reported associations with amyloid and tau burden, ⁴³ the cross‐sectional design precludes conclusions about directionality or causality. One cross‐sectional study ⁴⁰ found greater loneliness was related to accumulated tau pathology in the right entorhinal cortex and in the right fusiform gyrus. ³⁸ On the contrary, Wilson et al. ³⁴ analyzed autopsy data from older adults free from dementia at baseline with a subgroup developing AD during the follow‐up period. Although loneliness was related to dementia onset, no associations between baseline loneliness and the amount of beta‐amyloid plaque, tau tangle, or cerebral infarction were found. ³⁴ The authors concluded that lonely individuals might be more vulnerable to age‐related neuropathology, but not as a reaction to incident dementia or an early sign of its pathology. ³⁴ Van der Velpen et al. ⁵² reported loneliness was not associated with plasma neurodegeneration biomarkers (aβ40, aβ42, total tau, neurofilament light chain [NfL]) cross‐sectionally or longitudinally.

2. Brain structural differences (related to loneliness in cross‐sectional studies, mixed longitudinal evidence): Loneliness has been linked to structural brain abnormalities that may underlie CD, mostly supported by cross‐sectional evidence while longitudinal studies are less common and the results were mixed. A cross‐sectional study found that those with higher levels of loneliness had gray matter atrophy in the thalamus, hippocampus, and entorhinal cortex. ⁶⁵ Similarly, another cross‐sectional study reported higher levels of loneliness to be related to lower gray matter volume in the left amygdala/anterior hippocampus, the left posterior para‐hippocampus, and the left cerebellum. ⁶⁶ . Furthermore, an cross‐sectional investigation stratified by participants’ cognitive status (e.g., subjective cognitive decline [SCD], MCI) showed that among individuals with SCD, loneliness was related to smaller regional grey matter volume in the bilateral thalamus; and among individuals with MCI, loneliness was related to decreased grey matter volume in the left middle occipital gyrus and the cerebellar vermal lobules I−V. ⁵³ Structural brain changes have also been tested as mediators for the association between loneliness and CD. Park et al. ³⁶ found the cross‐sectional association between loneliness and cognitive performance was mediated by the volume of WMHs among men only, yet the relationship was not mediated by hippocampal or gray matter volumes.

In terms of longitudinal evidence, Dabiri et al. ³⁰ reported grey matter volume mediated the longitudinal relationship between loneliness and CD. A longitudinal investigation found that higher levels of loneliness were related to elevated volumes of periventricular WMHs (PWMHs), deep WMHs (DWMHs), and total WMHs over 5 years. ⁶⁷ Nonethelesss, Sole‐Padulles et al. ⁴² reported loneliness was not associated with either baseline or change in hippocampal volume and cortical thickness over time.

1. Cardiovascular risks (mixed cross‐sectional and longitudinal findings): Cardiovascular indicators emerged as another physiological pathway through which loneliness may contribute to CD with mixed findings. For instance, in a cross‐sectional study, greater loneliness was associated with the presence of microinfarcts in post mortem brain tissues. ³⁵ However, another cross‐sectional study by Chen et al. ⁶⁸ did not find a correlation between loneliness and cerebral blood flow. Two longitudinal studies that tested cardiovascular disease risk factors (e.g., blood pressure, HbA1C, etc.) as potential mediators found no significant effects. ⁴¹ , ⁶²

2. Neurotrophic factors (moderation supported by one longitudinal study): A longitudinal study by Dabiri et al. ³⁰ measured brain‐derived neurotrophic factor (BDNF) levels with RNA sequencing of post mortem tissues extracted from the gray matter. They found that low BDNF amplified loneliness‐related decline in visuospatial and semantic memory, while high BDNF was protective.

3. AD‐related gene expression (related to loneliness according to longitudinal studies): Post mortem transcriptomic studies show that loneliness is associated with widespread differential gene expression in both the nucleus accumbens and dorsolateral prefrontal cortex, implicating pathways linked to AD, psychiatric disorders, and immune dysfunction. ³² , ³³

4. General health indicators (mediation supported by longitudinal studies): Functionality, self‐rated health, number of chronic conditions were significant mediators of the longitudinal association between loneliness and CD, suggesting general health could explain part of the relationship. ⁶⁹ , ⁷⁰

5. Inflammation (moderation supported by one longitudinal study, medication not supported by one longitudinal study): Higher IL‐6 levels intensified loneliness‐related decline over time, ³⁰ while CRP was not supported as a mediator of the longitudinal association between loneliness and CD. ⁶⁹

6. HPA‐axis functioning (mediation supported by one cross‐sectional study): HPA‐axis functioning has also been supported as an underlying pathway, with one study demonstrating that bedtime cortisol levels mediated the cross‐sectional relationship between loneliness and worse cognitive performance. ⁷¹

3.3.2. Psychobehavioral pathways

1. Psychological pathways (depressive symptoms supported by longitudinal evidence, sense of control, perceived constraint, neuroticism supported by one cross‐sectional study): Several psychobehavioral pathways emerged as potential mechanisms linking loneliness to CD. Depressive symptoms were the most frequently examined psychological pathway, and findings were consistent across the studies. All four longitudinal studies supported that depressive symptoms mediated the association between loneliness and CD in older adults. ²⁹ , ³⁰ , ³¹ , ⁴³ Other psychological mediators included sense of control and neuroticism. ⁷² , ⁷³ Lower sense of control and higher perceived constraint separately mediated the cross‐sectional association of loneliness and CD. ⁷³ Additionally, loneliness was cross‐sectionally related to worse cognitive functioning through an indirect pathway of higher neuroticism. ⁷² Factors that were not supported as mediators were anxiety (longitudinal) ⁴³ and a sense of personal mastery (cross‐sectional). ⁷³

2. Behavioral pathways (leisure activities supported by one cross‐sectional study, social participation supported by one longitudinal study): In terms of behavioral pathways, the cross‐sectional association between loneliness and cognitive function was mediated by reduced leisure activities, including physical, social, and mental activities. ⁷⁴ Social participation, measured at 5‐year follow‐up by frequency of attending organized meetings, attending religious services, volunteering, and socializing with family and friends, mediated the association between baseline loneliness and cognitive functioning in 10‐year follow‐up. ⁴⁷

3.4. Social isolation and cognitive decline

3.4.1. Physiological pathways

Three cross‐sectional and five longitudinal studies investigated physiological pathways linking social isolation to CD. The pathways investigated included (1) AD pathology, (2) neurotrophic factors, (3) brain structural differences, (4) cardiovascular risks, (5) inflammation, and (6) oral health. Overall findings included a mix of positive and null associations.

1. AD pathology (moderation supported by one longitudinal study): Social isolation was not directly related to Aβ deposition measured with PET scan, but baseline social isolation was associated with greater amyloid‐β‐related CD over 3 years. ³⁹ Nonetheless, baseline Aβ was not related to a longitudinal change in social engagement. ³⁹

2. Neurotrophic factors (cross‐sectionally related to social isolation according to one study): One mechanism through which social isolation could be related to accelerated CD is through lower levels of neurotrophic factors. Social isolation was cross‐sectionally related to lower serum BDNF levels. ⁵⁶

3. Brain structural differences (association with social isolation supported by both cross‐sectional and longitudinal evidence, but longitudinal mediation was not supported by one study): Brain structural differences as physiological pathways for the association between social isolation and CD were examined in three studies (one cross‐sectional and two longitudinal). Cross‐sectionally, social isolation was associated to lower total brain volume and regional brain volumes in the temporal lobe, occipital lobe, cingulum, hippocampus, and amygdala, as well as higher white matter lesion volume, and depressive symptoms partially mediated these associations. ⁵⁹ A longitudinal investigation over 7 years found both baseline social isolation and the longitudinal increase in social isolation were related to smaller volumes of the hippocampus at the end of the study and reduced cortical thickness compared to baseline. ⁴⁸ However, hippocampal volume did not mediate the relationship between social isolation and global cognition. ⁴⁸ Social isolation was associated with a greater rate of hippocampal atrophy over four years, and this effect remained significant after controlling for loneliness. ⁴⁴

4. Cardiovascular risks (related to social isolation according to one cross‐sectional study): A cross‐sectional study provided evidence linking social isolation to cardiovascular risks for dementia. ⁵⁸ Cerebral vasodilatory response (CVR) is the ability of blood vessels to dilate and constrict in response to carbon dioxide, which is critical for maintaining adequate blood supply to the brain. ⁵⁸ Older adults with smaller social network size exhibited lower CVR in in the hippocampus, and less social network complexity was negatively associated with CVR in the regions previously implicated in social cognition, including hippocampus, right amygdala, right entorhinal cortex, and right para‐hippocampal gyrus. ⁵⁸

5. Inflammation (one cross‐sectional study supported mediation among men only): One cross‐sectional study reported sex differences in the association between social isolation and inflammation in relation to cognitive functioning. Social isolation was related to higher levels of inflammation (CRP and fibrinogen) in older men but not older women. ⁵⁵ Among older men, CRP and fibrinogen separately partially mediated the association between social isolation with cognition, yet albumin did not mediate the relationship. ⁵⁵

6. Oral health (mediation supported by one longitudinal study): Poor oral health, measured by number of missing teeth, mediated the association between baseline social isolation and subsequent cognitive performance in 4 years. ⁴⁵

3.4.2. Psychobehavioral pathways

1. Psychological pathways (depressive symptoms supported by cross‐sectional and longitudinal evidence, loneliness supported by cross‐sectional evidence, psychological well‐being supported by one longitudinal study): Depressive symptoms and loneliness are the most examined psychological mediators of the relationship between social isolation and CD with consistent evidence. ³¹ , ³⁷ , ⁵⁴ , ⁷⁵ Loneliness was cross‐sectionally supported as a mediator between the association of social isolation and CD. ³⁷ , ⁷⁵ One longitudinal study found that depressive symptoms independently mediated the associations of older adults’ social isolation and incident dementia over a 15‐year period. ³¹ Psychological well‐being, encompassing both positive and negative affect as well as personality traits, has also been identified as a mediator in this longitudinal association. ⁴¹

2. Behavioral pathways (sleep disturbance and physical activity supported by longitudinal evidence): Increased sleep disturbance ⁵⁷ mediated the longitudinal association between social isolation and CD over 8 years, and reduced physical activity ⁴⁶ mediated the association over 3 years.

3.5. Risk of bias

Results of the risk of bias assessment are summarized in Table 5 for cross‐sectional studies and in Table 6 for longitudinal studies. Among cross‐sectional studies, methodological issues identified include inclusion criteria not clearly defined, and study subject and research setting not described in detail, which could introduce selection bias and limit the external validity of the research findings. Measurement concerns for the exposure variables (i.e., social isolation and loneliness) exist for both cross‐sectional and longitudinal studies. For example, three studies relied on single‐item measures of loneliness, which are generally less reliable than multi‐item scales. ⁷⁶ Moreover, the operationalization and measurement of social isolation were even more heterogeneous than those of loneliness, which might make the findings less comparable across studies. Seven out of the total 19 longitudinal studies had incomplete follow‐up yet the reasons for loss to follow‐up were not explored and described. Additionally, nine out of 19 longitudinal studies did not describe strategies to address incomplete follow‐up.

TABLE 5.

Results of the quality assessment using the JBI risk of bias checklist for cross‐sectional studies.

Cross‐sectional studies (n = 18)
Author (year)	Were the criteria for inclusion in the sample clearly defined?	Were the study subjects and the setting described in detail?	Was the exposure measured in a valid and reliable way?	Were objective, standard criteria used for measurement of the condition?	Were confounding factors identified?	Were strategies to deal with confounding factors stated?	Were the outcomes measured in a valid and reliable way?	Was appropriate statistical analysis used?
Hirabayashi et al. (2023) 59	Yes	Yes	Unclear	NA	Yes	Yes	Yes	Yes
Salinas et al. (2017) 56	No	Yes	Yes	NA	Yes	Yes	Yes	Unclear
Imai et al. (2022) 65	Yes	Yes	Yes	NA	Yes	Yes	Yes	Yes
Canli et al. (2017) 33	Yes	Yes	Unclear	Yes	No	NA	Yes	No
Montoliu et al. 71	Yes	Yes	Yes	NA	Yes	Yes	Yes	Yes
Jang et al. (2021) 75	No	Yes	Yes	NA	Yes	Yes	Yes	Yes
Chen et al. (2022) 68	No	Yes	Yes	NA	Yes	Yes	Yes	Yes
Qi et al. (2023) 55	Yes	Yes	Yes	NA	Yes	Yes	Yes	Yes
Yang et al. (2021) 73	No	No	Yes	No	Yes	Yes	No	No
Foong et al. (2018) 72	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Duzel et al. (2019) 66	Yes	No	Yes	NA	Yes	Yes	Yes	Yes
Uquillas et al. (2018) 38	Yes	No	Yes	NA	Yes	Yes	Yes	Yes
Yang et al. (2020) 37	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Park et al. (2023) 36	Yes	No	Yes	Yes	Yes	Yes	Yes	Yes
Lao et al. (2024) 35	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Du et al. (2024) 74	Yes	Yes	Yes	NA	Yes	Yes	Yes	Yes
Chang et al. (2024) 58	Yes	Yes	Yes	NA	Yes	Yes	Yes	Yes
Bai et al. (2024) 54	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

Abbreviation: JBI, Joanna Briggs Institute.

TABLE 6.

Results of the quality assessment using the JBI risk of bias checklist for longitudinal studies.

Longitudinal studies (n = 19)
Author (year)	Was the exposure measured in a valid and reliable way?	Were confounding factors identified?	Were strategies to deal with confounding factors stated?	Were the groups/participants free of the outcome at the start of the study (or at the moment of exposure)?	Were the outcomes measured in a valid and reliable way?	Was the follow up time reported and long enough for outcomes to occur?	Was follow up complete, and if not, were the reasons to loss to follow up described and explored?	Were strategies to address incomplete follow up utilized?	Was appropriate statistical analysis used?
Qi et al. (2023) 57	Yes	Yes	Yes	NA	Yes	NA	Yes	Yes	Yes
Lammer et al. (2023) 48	Yes	Yes	Yes	Yes	Yes	Unclear	No	No	Yes
Zhou et al. (2020) 41	Unclear	Yes	Yes	Yes	Yes	No	Yes	No	Yes
Canli et al. (2018) 32	Unclear	Yes	Yes	Yes	Yes	Unclear	Unclear	Unclear	Yes
Duan et al. (2017) 67	Yes	Yes	Yes	Yes	Yes	Unclear	Yes	No	Yes
Wilson et al. (2007) 34	Unclear	Yes	Yes	Yes	Yes	Yes	Yes	NA	Yes
Zhang et al. (2022) 53	No	Yes	Yes	No	Yes	Yes	Unclear	Unclear	Yes
Yu and Ng (2023) 69	Yes	Yes	Yes	Yes	Yes	Unclear	Yes	Yes	Yes
Power et al. (2020) 43	Yes	Yes	Yes	Yes	Yes	Unclear	Yes	No	Yes
Kim et al. (2020) 29	Yes	Yes	Yes	Unclear	Yes	Unclear	No	Yes	Yes
Zhong et al. (2017) 50	No	Yes	Yes	NA	Yes	Yes	No	Yes	Yes
Sole‐Padulles et al. (2022) 42	Unclear	Yes	Yes	Yes	Yes	Unclear	Unclear	No	Yes
Biddle et al. (2019) 77	Yes	Yes	Yes	Yes	Yes	Unclear	No	Unclear	Yes
Murayama et al. (2024) 44	Yes	Yes	Yes	Unclear	Yes	Unclear	Unclear	Unclear	Yes
Duffner et al. (2024) 31	Yes	Yes	Yes	Yes	Yes	Yes	No	No	Yes
Van der Velpen et al. (2024) 52	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	Yes
Khalaila et al. (2024) 45	Yes	Yes	Yes	Yes	Yes	Yes	No	No	Yes
Hopper et al. (2024) 46	Yes	Yes	Yes	Yes	Yes	Unclear	Yes	Yes	Yes
Dabiri et al. (2024) 30	Unclear	Yes	Yes	Yes	Yes	Yes	No	No	Yes

Abbreviation: JBI, Joanna Briggs Institute.

4. DISCUSSION

Empirical evidence shows that higher levels of loneliness and social isolation are associated with CD among older adults; however, questions remain regarding the mechanisms through which loneliness and social isolation affect CD. This review synthesized evidence on how physiological and psychobehavioral mechanisms may explain the associations of loneliness and social isolation with CD in later life. Drawing on 37 empirical studies with a total of 185,794 participants, we identified multiple physiological and psychobehavioral pathways connecting loneliness and social isolation to CD. The literature has focused more extensively on mechanisms linking loneliness with CD, with comparatively fewer studies examining mechanisms linking social isolation to CD. Although there was considerable overlap in the mechanistic pathways identified for loneliness and social isolation, distinctions also emerged. Similar categories of physiological pathway domains (e.g., brain structural change) were identified for both loneliness and social isolation. However, their relationships with specific indicators (e.g., total gray matter volume) differ. Additionally, more behavioral pathways were examined for social isolation while more psychological pathways identified for loneliness. Below, we provide our interpretations of existing evidence on physiological pathways and psychobehavioral pathways, followed by evaluation of the strength and weakness of existing literature, future research directions, and strength and weakness of the current review.

4.1. Physiological pathways

4.1.1. Risk factor versus prodromal hypothesis

A debate in the literature concerns the nature of the association between loneliness, social isolation, and CD: are loneliness and social isolation risk factors that precede and contribute to cognitive deterioration, or are they early signs/symptoms that an individual may be developing dementia (i.e., the prodromal hypothesis)? The current review found evidence supporting both sides of this debate. Overall, more studies have examined loneliness and isolation as risk factors, with multiple longitudinal investigations supporting their role in predicting both CD and dementia onset. ³⁰ , ³¹ , ⁵² , ⁵⁷ The cognitive vulnerability hypothesis posits that loneliness and social isolation erode cognitive reserve—the brain's capacity to maintain function despite age‐related changes—making individuals more susceptible to neuropathological damage. ³⁴ This systematic review identified several mechanisms that may underlie this vulnerability. Both loneliness and social isolation may activate the HPA axis and sympathetic nervous system, which in turn may contribute to vascular dysregulation, immune dysfunction, and neurodegeneration. ⁷¹ Molecular evidence from two longitudinal studies suggests that loneliness is linked to altered expression of Alzheimer's‐related gene sets, ³² , ³³ offering additional information for increased vulnerability among lonely individuals. Moreover, findings support that better social connection with others preserve cerebrovascular health and buffer against cognitive deterioration. ⁵⁸

Additionally, some research has found important bidirectional relationships between social isolation, loneliness, and CD. As an example, a longitudinal study examining social engagement and CD found partial evidence that a bidirectional relationship may exist. ⁷⁷ This study found a significant interaction effect between social engagement and Aβ; adults with greater baseline Aβ levels and lower baseline social engagement had greater CD three years later. ⁷⁷ Moreover, this study found baseline CD scores predicted social engagement three years later, illustrating the potential for a bidirectional relationship between social engagement and CD. ⁷⁷ More empirical investigations are necessary to disentangle the directionality of the association.

4.2. Psychobehavioral pathways

Psychobehavioral pathways linking loneliness and social isolation show commonalities but also important distinctions. For loneliness, depressive symptoms are the most consistently supported mediator across longitudinal studies, ³¹ , ³⁷ , ⁵⁴ , ⁷⁵ with additional contributions from lower sense of control, ⁷³ higher neuroticism, ⁷² and reduced engagement in leisure and social activities. ⁷⁴ Although loneliness was associated with symptoms of anxiety, it was not supported as a mediating pathway and the researchers suggested that anxiety can have both positive and negative effects on cognition, as it can increase attention. ⁴³ Core features of neuroticism—such as heightened fear of social rejection and low self‐esteem—may foster chronic stress and reduce opportunities for cognitively stimulating social interactions. ⁷² , ⁷⁸ These findings suggest that loneliness may undermine cognitive resilience by intensifying negative psychological states and reducing participation in cognitively enriching activities, both of which have been linked to CD in prior research. In contrast, studies on social isolation highlight a wider range of behavioral mechanisms, including sleep disturbance ⁵⁷ and reduced physical activity. ⁴⁶ With the existing evidence, it remains unclear whether these differences reflect distinct underlying mechanisms or variations in what prior studies chose to test. Future studies are needed to explore whether behavioral pathways identified for social isolation, such as sleep disturbance and physical inactivity, might also play a role in the association between loneliness and CD.

4.3. Appraisal of current evidence and future directions

The breadth of recent research on physiological and psychobehavioral mechanisms is noteworthy; such a wide range of evidence on the pathways linking loneliness, social isolation, and CD did not exist just a few years ago. This growth reflects an exciting shift in the field, but it also reveals how much more remains to be understood. For example, the temporal dynamics of mediating pathways remain unclear: what level or duration of loneliness and social isolation triggers physiological or behavioral changes, and over what timeframe might it lead to cognitive change? Another important question is whether reversing loneliness or social isolation can also reverse the associated physiological and psychobehavioral changes, ultimately leading to preserved cognitive function. Some of these questions could be addressed through randomized controlled trials aimed at reducing loneliness or social isolation. For example, one clinical trial that reduced social isolation by providing frequent, engaging conversational interactions via the Internet showed that an increased amount of social interaction led to greater dorsal attention network connectivity, using resting‐state functional magnetic resonance imaging (fMRI). ⁷⁹ The study also showed that the intervention led to increased social engagement, ⁷⁹ , ⁸⁰ suggesting that reducing social isolation even in later life could modify physiological and psychobehavioral factors. Animal model studies can also help posit hypotheses on biological sequences. For instance, in mice model studies, social isolation was linked to HPA‐axis dysregulation that decreased BDNF and further neurodegeneration. ²⁵ Additionally, future human studies using time‐sensitive measures of social isolation and loneliness, ⁸¹ , ⁸² coupled with longitudinal biomarkers and behavioral assessments, are needed to investigate the temporal dynamics of mediating pathways. Furthermore, because most pathways have been examined in only two or three studies with different methods, conducting a meta‐analysis to estimate direct and indirect effect sizes is not yet feasible nor informative. More rigorous investigations are needed to enable meta‐analysis and establish pulled effect size of indirect pathways. Building evidence on the effect size of the pathways can contribute to a clearer understanding of their magnitude of influence on cognition, which can inform intervention targets.

An interesting pattern emerged in our review: more psychological pathways (e.g., stress and affective processes) were reported for loneliness, whereas more behavioral pathways (e.g., physical activity, social engagement, and health behaviors) were identified for social isolation. This distinction may reflect fundamental differences in the constructs themselves. Loneliness is a subjective, appraisal‐based experience that can arise even in the presence of social contact and may therefore operate primarily through emotional and stress‐related mechanisms. By contrast, social isolation reflects an objective lack of social ties or interaction, which may more directly constrain opportunities for cognitively and physically stimulating activities, aligning more with behavioral pathways. At the same time, these observed patterns could reflect differences in what prior studies have chosen to measure or emphasize. Future work is needed to jointly assess loneliness and social isolation, including their concordance and discordance (often conceptualized as social asymmetry) and to test multiple mechanistic pathways within the same study. Recent research has begun to examine how mismatches between subjective and objective social connectedness relate to health outcomes, ⁸³ highlighting the importance of explicitly modeling these distinctions to determine whether observed differences in mechanisms reflect genuine underlying processes or methodological artifacts.

Several methodological concerns in the existing literature may increase the risk of bias, including heterogeneity in how loneliness and social isolation are defined and measured, and selection bias that might be introduced with convenience sampling. For loneliness, studies varied widely in their operationalization. Some adopted categorical cut‐offs to distinguish between “no loneliness” and “severe loneliness,” while others analyzed loneliness as a continuous construct. Such differences complicate direct comparisons across studies and may partly account for inconsistent findings. Moreover, few studies examined the duration of loneliness and social isolation, even though short‐term versus persistent experiences may have distinct implications for cognitive outcomes. Longitudinal analyses that adopt a measurement burst design to capture duration and intraindividual change within days, weeks, and months are needed to address this gap. ⁸⁴ Figure 2 outlines the empirically examined pathways, yet it is worth noticing that not all pathways are examined with equal quality of evidence: a few studies had a small convenience sample of less than 70 participants, ³³ , ⁶⁸ , ⁷³ and the results are more likely to be influenced by sampling bias. Racial/ethnic minority and individuals with lower socioeconomic status, although are at higher risks of social isolation, loneliness, and dementia, ⁸⁵ , ⁸⁶ , ⁸⁷ are underrepresented in the current literature. A combined research approach on how social determinants (e.g., isolation due to socioeconomic status) may interact with biological vulnerabilities are needed, as such investigations might reveal more contextualized risks for CD and identify for which populations certain mechanisms have more explanatory capability and offer insight into appropriate intervention approaches for different groups.

4.4. Potential mechanisms beyond current evidence

In addition to the mechanisms empirically examined in the literature, several other pathways may be relevant but remain untested to our knowledge. One example is gut–brain axis dysregulation, which has been proposed as a therapeutic target for dementia prevention and treatment and may relate to social isolation and loneliness. ⁸⁸ , ⁸⁹ Microbiomes can be shared through social contact while socially isolated individuals have fewer opportunities for such exchange. ⁸⁸ , ⁸⁹ Gut microbiota dysbiosis is thought to promote CD in older adults by increasing intestinal permeability, allowing pro‐inflammatory microbial products (e.g., lipopolysaccharides and bacterial amyloids) to enter systemic circulation, disrupt the blood–brain barrier, and potentially leading to increased neuroinflammation. ⁹⁰ , ⁹¹ Second, dietary patterns among older adults who experience social isolation and loneliness also warrant consideration. A recent scoping review suggests that loneliness and social isolation are associated with less healthy eating behaviors, including reduced fruit and vegetable intake and overall poorer diet quality. ⁹² These behaviors have been linked with lower intake of micronutrients such as magnesium, potassium, vitamin B6, folate, and vitamin C, ⁹³ which in turn have been associated with CD and dementia onset. ⁹⁴ Third, mindfulness, the practice of being fully present and aware of one's current experience, also represents a potentially relevant pathway. Evidence indicates that mindfulness training and practices can reduce loneliness and increase social contact among older adults. ⁹⁵ , ⁹⁶ However, we did not identify empirical studies that examine mindfulness as an underlying pathway for the relationships between loneliness, social isolation, and cognitive outcomes in older adults.

Across all three proposed pathways—gut–brain axis disruption, diet, and mindfulness—the inferential chains presented rely on evidence that differs in directness, population characteristics, and methodological design at each step. These mechanisms are best understood as theoretical hypotheses intended to motivate future research. Considering these additional pathways highlights the opportunities for future work to further clarify underlying mechanisms.

4.5. Suggestion for intervention programs

Findings from this review point to several mechanistic targets that may be leveraged in non‐pharmacological interventions (Figure 2). Future programs aimed at reducing social isolation and loneliness to delay age‐related CD could investigate whether they impact physiological and psychobehavioral factors summarized here. For example, the intervention behavioral mechanisms could be evaluated by assessing sense of control, sleep quality, enhanced social participation, and physical activities. The intervention outcomes can also be physiological pathways identified (e.g., neuroimaging, blood‐based AD biomarkers, inflammatory markers, and neurotrophic factors). These trials could generate valuable evidence on the modifiability of these mechanisms and accelerate clinical translation.

4.6. Limitations of the current review

This review comprehensively synthesized evidence on the physiological and psychobehavioral mechanisms and clarified how loneliness and social isolation can be linked to CD in later life. A few limitations of the current review need to be acknowledged when interpreting the findings. All included studies were observational and therefore limited in the ability to establish causality. The inclusion criteria required loneliness or social isolation to be treated as the exposure variable, which means studies examining CD as a precipitant of social withdrawal were not captured in this review. Given the possibility that loneliness may serve as a prodromal symptom of dementia, future reviews incorporating both directions of the relationship—ideally using studies with bidirectional designs such as cross‐lagged panel models—would provide a more complete picture of this complex association. Furthermore, conducting a meta‐analysis to quantify the strength of the indirect effect of the identified mechanisms was not feasible given the lack of comparable studies that examined similar mechanisms.

To capture a wide range of relevant evidence, we did not require the physiological pathways to have a mediating effect. However, having a physiological outcome with direct dementia implications does not guarantee mediating relationships. For instance, although loneliness is linked to upregulated AD‐related gene expression and such gene expression is associated with higher likelihood of CD due to AD risks ³² , ³³ —it does not mean that loneliness can lead to CD through AD‐related gene expression as a pathway. That requires the indirect effect to be significant. Although we did identify a few empirical studies that analyzed physiological pathways as a mediator (see Figure 2), future studies are needed to further examine these potential physiological pathways identified in this current review are true mechanisms. More intentional examinations of physiological biomarkers (e.g., gray matter volume and microinfarcts) as mediators for the association between social isolation, loneliness, and cognition can help provide empirical support or falsify the hypothesis. In examining the behavioral pathways which examine how loneliness is associated with CD, some of these studies used social participation in activities as the key mediating variable. ²⁹ , ⁷⁴ Although social participation was operationalized behaviorally in these two studies, we acknowledge that social participation instruments vary across the literature and can capture subjective dimensions of felt connection that overlap conceptually with loneliness; future studies examining this psychobehavioral pathway should explicitly employ frequency‐based or activity‐based measures of social participation to minimize mediator‐exposure confounding.

5. CONCLUSION

This review is among the first to integrate evidence on physiological and psychobehavioral pathways linking loneliness and social isolation to CD. Physiological mechanisms identified include AD pathology, inflammation, cardiovascular risks, neurotrophic factors, AD‐related gene expression, HPA‐axis functioning, oral health, and general health, though findings remain mixed and the temporal sequence of these changes is not yet established. Psychobehavioral pathways, the most examined was depressive symptoms, along with neuroticism, sense of control, sleep disturbance, and reduced engagement in social and leisure activities, mediated the associations between social isolation and loneliness with CD. Future longitudinal and biomarker‐informed studies are needed to clarify directionality and guide intervention trials to prevent age‐related abnormal CD.

AUTHOR CONTRIBUTIONS

Kexin Yu, Karra D. Harrington, and Hiroko H. Dodge contributed to the design and conception of the study. Kexin Yu, Karina Van Bogart, Jee Eun Kang, and Harry O. Taylor conducted screening, data extraction, and risk of bias evaluation. Kexin Yu and Karina Van Bogart drafted and edited the full manuscript. Hiroko H. Dodge, Lisa C. Silbert, and Ihab Hajjar provided critical feedback throughout the review and manuscript preparation processes that informed the direction of the investigation. All authors contributed to manuscript revision, read, and approved the submitted version.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest. Author disclosures are available in the Supporting Information.

Supporting information

Supporting Information: alz71585‐sup‐0001‐ICMJE.pdf

Supporting Information: alz71585‐sup‐0002‐Table‐1.docx

Supporting Information: alz71585‐sup‐0003‐Table‐2‐5.xlsx