Table 1.
Overview of literature on the association between social network characteristics and type 2 diabetes risk
| Study | Study name and population | Follow-up duration (years) | N | (Incident) T2D cases (n) | Mean age or age range (years) | Measures of social network | Measurement of diabetes | Main results | Study limitations |
|---|---|---|---|---|---|---|---|---|---|
| Cross-sectional study design | |||||||||
| Brinkhues (2017) [11] |
The Maastricht Study. Population-based, oversampled diabetes. |
NA | 2861 | 808 | 40–75 | Various structural and functional network characteristics | OGTT |
Small network size associated with newly and previously diagnosed diabetes. In women, larger distance, more household members and fewer friends associated with newly diagnosed and known diabetes. Lack of social participation associated with diabetes in men and women. Living alone associated with diabetes in men only. Less support in men and women with new and previous diabetes vs those with NGT. |
Relatively healthy population |
| Jones (2016) [35] |
NHATS. Medicare representative survey, USA. |
NA | 6942 | Unknown | 65–94 | Number of important people (0–5) | Registry | Diabetes associated with smaller network. | Study focused on multimorbidity; not diabetes specific |
| Hempler (2013) [36] |
Part of Danish health and morbidity survey. Diabetes clinic vs web panel, diabetes vs general population. |
NA | Diabetes clinic: 1084; web panel: 1491; general population: 15,165 | 2575 | Diabetes clinic: 64; web panel: 60; general population: 50 | Social structure (household, contact family/friends) and function (count on others help) | Clinic or self-report | Living alone (diabetes risk: OR 1.75–1.64), meeting family less than once per month (diabetes risk: OR 1.78–2.35). | Limited assessment of social network |
| Longitudinal study design | |||||||||
| Hendryx (2020) [18] |
Women’s Health Initiative. Postmenopausal women from 40 US clinical centres. |
14 | 139,924 | 19,240 | 50–79 | Social support (9 item scale), social network (range: 0–5), social strain (4 item scale) | Self-reported new diagnosis of T2D treated with insulin or oral drugs |
Social support is protective for incident diabetes. BMI, PA and depression mediate association between support and incident diabetes. |
Limited assessment of social network |
| Lukaschek (2017) [12] |
MONICA/KORA. General population. |
14 | 6839 | 551 | 25–74 | SNS (one item), living alone, SNI (dichotomous) | Self-reported or validated by physicians | Low SNS associated with doubled diabetes risk, partly via social isolation and living alone, but only in men (living alone in men, HR 1.62 [95% CI 1.22, 2.15]). | A construct of social network (SNI) was used; limited assessment of social network |
| Laursen (2017) [19] |
ELSA. General population. |
7.7 | 7662 | 804 | 50–91 | Social relationships (spouse, children, other immediate family, friends; score 0–4), social support per relationship (score 0–36), social strain (3 item) | Self-reported or screen-detected T2D | Higher support and low strain are associated with lower diabetes risk, dependent on BMI and health behaviour (smoking, alcohol consumption and PA). | Limited assessment of social network |
| Altevers (2016) [20] |
MONICA/KORA. General population. |
15.5 | 8952 | 904 | 30–74 | SNI (dichotomous) | Self-reported T2D validated with physicians’ records | Poor vs good structural support is associated with increased diabetes risk in men and women, and only in men after adjustment (age, survey, smoking, alcohol, PA, parent with diabetes, BMI, hypertension, dyslipidaemia) (men, HR 1.31 [95% CI 1.11, 1.55]; women HR 1.10 [95% CI 0.88, 1.37). | A construct of social network (SNI) was used; limited assessment of social network |
| Hilding (2015) [13] |
Stockholm diabetes prevention programme. General population. |
8–10 | 4963 | 143 | 35–56 | AVSI index (6 items; score 0–12), civil status (partner, yes/no), participation in social activity (2 items; dichotomous) | OGTT |
High AVSI is associated with low diabetes risk in women, partly dependent on BMI, PA and smoking. In men, high AVSI is associated with high diabetes risk. Social participation decreases diabetes risk in men and women. Having a partner is protective for diabetes in men only: for T2D, OR 0.57 (95% CI 0.33, 0.97) and OR 0.61 (95% CI 0.34, 1.08) in age- and multi-adjusted models, respectively. |
A construct of social network (AVSI) was used; Limited assessment of social network |
| Meisinger (2009) [14] |
MONICA/KORA Augsburg Cohort Study. General population. |
10.9 | 8804 | 673 | 35–74 | Living alone assessed by one question | Questionnaire and clinical validation | Living alone is a risk factor for diabetes in men, not women: HR 1.66 (95% CI 118, 2.34) vs HR 0.86 (95% CI 0.58, 1.26). | Limited assessment of social network (only living alone) |
| Strodl (2006) [15] |
Australian women’s health survey. Women from the general population. |
3 | 10,300 | 243 | 70–74 | DSSI (11 items; 2 factors: social satisfaction and social interaction), marital status | Participants asked, ‘did a doctor inform you that you have diabetes?’ (no distinction between T1D or T2D) | Moderate social support predicts new diabetes in elderly women in univariate analysis. | A construct of social network (DSSI) was used; Limited assessment of social network |
| Lidfeldt (2005) [16] |
Study name: NA. Women with IGT from population registry. |
2.5 | 461 | 55 | 50–64 | Living alone, marital status | OGTT to assess T2D | Living alone is a risk factor for T2D (age-adjusted OR 2.47 [95% CI 1.06, 5.71]) but depends on smoking status (fully adjusted OR 2.07 [95% CI 0.62, 6.88]). | Limited assessment of social network (only living alone and marital status) |
| Kumari (2004) [17] |
Whitehall II study. Occupational cohort study, London (UK). |
10.5 (4 years follow-up momentsa) | 10,308 | 361 | 35–55 | Social support (Close Persons Questionnaire), questions derived by Berkman and Syme. The network measures were summarised on a scale measuring the network beyond the household. | Self-reported, physician-diagnosed or OGTT | Effort–reward imbalance is associated with incident diabetes in men. | A construct of social network (Close Persons Questionnaire) was used; Limited assessment of social network |
| Norberg (2007) [37] |
Health survey. Occupational population, Umea (Sweden); age- and sex-matched nested case-referent study |
5.4 ± 2.6 | 584 | 191 | Cohort 1: 40; cohort 2: 50; cohort 3: 60 | Social network and emotional support (interview schedule for social interaction), social integration (AVSI, 7 questions), attachment (AVAT, 7 questions) | Medical registry-based diagnosis of T2D | Low AVAT associated with incident T2D in women (weak AVAT: OR 3.0 [95% CI 1.3, 7.0]). | A construct of social network (AVSI, AVAT) was used; Limited assessment of social network |
| Hill (2014) [38] |
Health and retirement study, USA Representative sample of ageing population |
4 | 5681 | 430 | 68 | Social relationships (spouse, children, other immediate family, friends; score 0–4), social support per relationship (score 0–36), social strain (3 items) | Self-reported | Negative friend support increases the risk for diabetes (OR 1.31 [95% CI 1.07, 1.62]). | Limited assessment of social network |
aFollow-up data were collected at four different points in time and these data were merged (no continuous follow-up)
AVAT, availability of attachment; AVSI, availability of social integration; DSSI, Duke social support index; IGT, impaired glucose tolerance; NA, not applicable; NGT, normal glucose tolerance; NHATS, National Health and Aging Trends Study; PA, physical activity; SNI, Social Network Index; T1D, type 1 diabetes; T2D, type 2 diabetes