Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
EPIDEMIOLOGY/GENETICS
Objective: To investigate the effect of five dimensions of social relationships on obesity and potential sex
differences in these associations.
Design and Methods: This study used longitudinal data from the Swedish Level of Living Surveys (LNU)
in 1991 and 2000. The sample included 3,586 individuals. The dimensions of social relationships
examined in this study include emotional support, frequency of visiting friends, marital status, marital
status changes, and a Social Relationships Index (SRI). Obesity status was based on BMI (kg/m2) and
calculated with self-reported measurements. The association between social relationships and the
incidence of obesity after 9 years of follow-up was evaluated through Poisson regressions.
Results: After controlling for confounders, we found that the lack of emotional support (RR ¼ 1.98; 95%
CI, 1.1-4.6) influenced the incidence of obesity among men. In addition, men with the lowest levels of
SRI (RR ¼ 2.22; 95% CI, 1.1-4.4) had an increased risk of being obese. Among women, SRI was not
significantly associated with obesity. Women who changed their marital status from married to unmarried
had lower risk of obesity (RR ¼ 0.39; 95% CI, 0.2-0.9).
Conclusions: This study provides evidence for the effect of social relationships on the incidence of
obesity, with significant differences by sex.
1
Department of Epidemiology, Institute of Social Medicine, Rio de Janeiro State University, Rio de Janeiro, RJ 20550-900, Brazil. Correspondence: Aldair J.
Oliveira (oliveira.jose.aldair@gmail.com) 2 Centre for Health Equity Studies (CHESS), Stockholm University/Karolinska Institutet, Sweden
suggest that internalized obligations to—but mainly the overt influ- tion to correct for self-reported bias because individuals tend to
ences of—network members tend to discourage poor health behav- underestimate their weights and overestimate their heights (24).
iors and encourage good ones (15). By this account, individuals who The following algorithms were adjusted for age and calculated sep-
do not regularly meet with friends have no social barriers to hinder arately by sex:
them from engaging in or maintaining poor eating, drinking, and
exercise habits, all of which contribute to obesity. A lack of contact For men : BMIC ¼ 0:202 þ 1:005 BMISR þ 0:014 age
with friends could decrease exposure to social control, which pre-
For women : BMIC ¼ 0:713 þ 1:023 BMISR þ 0:019 age
vents people from adopting a healthy lifestyle that incorporates
moderate or vigorous physical activity and healthy food consump- where BMISR is the value of the self-reported BMI, and age was in
tion. In relation to marital status, our hypothesis is that nonmarried years. BMIC is the new value of BMI. Aiming to assess the inci-
persons have less social support that encourages healthy lifestyles dence of obesity, a dichotomous variable was created based on LNU
compared with their married counterparts, which could lead to obe- 2000 information, where individuals with BMI < 30 kg/m2 were
sity. Moreover, nonmarried individuals are less likely exposed to classified as nonobese and individuals with BMI 30.0 kg/m2 were
forms of informal social control over deviant health behaviors due considered obese.
to the lack of a stable partner. Finally, poorer psychological well-
being and feelings of loneliness among the nonmarried could also
lead to poorer eating and exercise habits (16). Social relationships. Considering that a single validated instru-
ment to evaluate the social relationships was not available in LNU,
The role of social networks and social support on health is well we decided to the use three questions designed to cover different
established in the literature, and studies have shown significant gen- dimensions of social relationships including emotional support, mari-
der differences in how social support operates in relation to health tal status, and the magnitude of social interaction, in order to study
(15,17-19). These studies have shown that contact with friends and associations with obesity. Additionally, the social relationship ques-
emotional support have more important consequences on women’s tions have been evaluated and results show good validity and reli-
than men’s health (20). Marriage, however, is more protective for ability (25).
men than for women (21). Considering that women’s support net-
works often include close friends and relatives as confidantes, Five different indicators of social relationships were evaluated based
whereas men typically name their wives as their main source of sup- on the following questions, where applicable. First, emotional support
port, marital disruption appears to be more detrimental for men’s is based on the question, Sometimes we need other people’s help and
health than for women’s health (22). In spite of the abundance of support. Do you have a family member or friend who helps out if you
research on gender differences in social capital and general health, need to talk to someone about personal problems? The respondents
the literature on sex differences in the association between social sup- provided a dichotomous answer to these questions by answering
port and the specific phenomenon of obesity is comparably scarce. ‘‘yes’’ or ‘‘no’’. Second, frequency of visiting friends is based on the
question, Do you engage in any of the following leisure activities
In general, our hypothesis is that poor social relationships influence (e.g., visit friends and acquaintances)? The possible responses
the risk of obesity among both women and men, although sex differen- include ‘‘yes, often,’’ ‘‘yes, sometimes,’’ and ‘‘no’’. All of the ques-
ces may exist depending on the dimension of social relationships. tions were asked in reference to the 12 months period prior to the
Therefore, this study has two aims: (1) to examine whether there is an survey. The same questions appeared in both survey waves (i.e.,
association between different dimensions of social relationships (e.g., 1991 and 2000). Third, marital status is measured using three catego-
emotional support, frequency of visiting friends and marital status) and ries: ‘‘married/cohabiting,’’ ‘‘divorced/widowed,’’ and ‘‘single.’’ To
the incidence of obesity, and (2) to investigate whether there are sex investigate potential modifications of social relationships on obesity,
differences in the influence of poor social relationships on obesity. we created two variables to represent changes in social relationships.
Fourth, changes in marital status between the two time points of our
study are captured using four categories: married in 1991 and in
2000, married in 1991 and not married in 2000, not married in 1991
Methods and Procedures and married in 2000, and not married in either 1991 or 2000. Finally,
Design and study population we created a Social Relationships Index (SRI), aiming to investigate
The Swedish Level of Living Survey (LNU) is based on face-to-face changes in social relationships during the follow-up period, as a sum-
interviews with a representative sample of the Swedish population mary measure of two dimensions of social relationships, including
aged 18-75 years. The survey has a panel design, but to maintain emotional support and frequency of visiting friends. For instance,
the survey’s representativeness of the Swedish population at any level 1 (the worst scenario) refers to individuals with poor emotional
given time, each survey wave also includes younger generations and support or low frequency of visiting friends at both survey waves
newly arrived immigrants. Furthermore, the LNU allows different (i.e., 1991 and 2000); and level 4 (the best scenario) refers to individ-
approaches with the use of its data. This study uses a longitudinal uals with moderate to high levels of emotional support and high or
design, including 3,586 individuals who participated in both the moderate frequency of visiting friends at both survey waves. Table 1
1991 and 2000 LNU waves in our analytical sample. presents all of the possible combinations regarding the SRI
classification.
Measures
Obesity. The BMI (kg/m2) was calculated based on self-reported Covariates
measurements of weight and height. It was corrected with algo- The covariates that we included in the analysis were BMI in 1991, age,
rithms by Nyholm and colleagues (23) using the Swedish popula- marital status, social class, leisure-time physical activity, alcohol
In the 9 year follow-up data, there were no significant associations was some evidence that a low frequency of visiting friends (Model 4:
between the various dimensions of social relationships and the devel- RR ¼ 1.81; 95% CI, 0.8-4.1) was associated with the risk of being
opment of obesity among women, with the exception of marital sta- obese. In general, the models presented in Table 3 did not show sig-
tus. Women who reported divorce or widowhood in the LNU 1991 nificantly different results across the various dimensions of social rela-
data had a higher risk of obesity in the 2000 data compared with their tionships. In addition, the multiplicity interaction tests between the
married or cohabiting counterparts (Model 4: RR ¼ 1.63; 95% CI, social relationships indicators and the health behavior variables in pre-
1.1-2.6). In addition, there was some evidence, although not signifi- dicting obesity incidence were not significant. The only exception
cant, of an inverse relationship between individuals that reported a was between marital status and smoking, where an interaction was
low frequency of visiting friends (Model 4: RR ¼ 0.31; 95% CI, 0.1- found between being single and having never smoked among women
2.2) and the incidence of obesity among women. Contrarily, the in predicting a greater risk of obesity (RR ¼ 0.65; 95% CI, 0.4-0.9).
adjusted models presented in Table 3 showed a significant association
between poor emotional support (Model 4: RR ¼ 1.98; 95% CI, 1.1- Regarding social relationships changes, Table 4 presents a signifi-
3.8) and the development of obesity among men. Furthermore, there cant inverse association between women who were married in 1991
a
The statistics presented are based on individuals who participated in both LNU waves (i.e., 1991 and 2000).
TABLE 3 Relative risks (RR) and confidence intervals (CI 95%) of the association between social relationships and incidence
of obesity using Poisson regressions
Men (n ¼ 1686)
Emotional support
Yes 1.00 1.00 1.00 1.00 1.00
No 2.14 (1.8–3.7) 1.95 (1.1–3.8) 2.02 (1.1–3.8) 2.01 (1.1–3.8) 1.98 (1.1–3.8)
Frequency of visiting friends
Frequently 1.00 1.00 1.00 1.00 1.00
Sometimes 0.89 (0.6–1.2) 0.93 (0.7–1.3) 0.93 (0.6–1.3) 0.92 (0.6–1.3) 0.93 (0.7–1.3)
No 2.05 (1.1–4.0) 1.68 (0.7–3.8) 1.81 (0.8–4.1) 1.80 (0.8–4.1) 1.81 (0.8–4.1)
Marital status
Married or cohabited 1.00 1.00 1.00 1.00 1.00
Divorced or widower 1.71 (1.1–2.8) 1.56 (0.9–2.9) 1.62 (0.9–3.1) 1.62 (0.9–3.1) 1.60 (0.8–3.0)
Single 0.79 (0.5–1.3) 0.80 (0.5–1.2) 0.82 (0.6–1.2) 0.82 (0.5–1.2) 0.81 (0.5–1.2)
The outcome was defined as a dummy variable representing two categories: nonobese (individuals that did not getting obese during the follow-up period the reference
group) and obesity (incidence cases in LNU 2000 BMI 30 kg/m2); exposures, the social relationships variables, and covariates were included based on LNU 1991 in-
formation. The Swedish Level of Living Survey (9 years of follow-up); Model 1: emotional isolation þ number of friends þ frequency of visiting friends þ marital status þ
age þ social class; Model 2: Model 1þ physical activity; Model 3: Model 2þ alcohol consumption; Model 4: Model 3þ smoking.
TABLE 4 Frequency, relative risks (RR) and confidence intervals (CI 95%) of the association between social relationships
changes and the incidence of obesity using Poisson regressionsa
Men (n ¼ 1686)
SRI
Level 4 1509 (8.2) 1.00 1.00 1.00 1.00 1.00
Level 3 80 (3.7) 0.46 (0.2–1.4) 0.46 (0.1–1.4) 0.44 (0.1–1.4) 0.45 (0.2–1.4) 0.45 (0.2–1.4)
Level 2 72 (16.4) 2.00 (1.1–3.6) 1.98 (1.1–3.6) 1.92 (1.1–3.5) 1.92 (1.1–3.5) 1.93 (1.1–3.5)
Level 1 36 (22.2) 2.70 (1.4–5.0) 2.34 (1.2–4.6) 2.19 (1.1–4.3) 2.20 (1.1–4.3) 2.22 (1.1–4.4)
Marital status changes
Marriage/marriage 1040 (7.9) 1.00 1.00 1.00 1.00 1.00
Marriage/non- marriage 109 (11.0) 1.37 (0.8–2.4) 1.24 (0.7–2.2) 1.29 (0.7–2.3) 1.29 (0.7–2.3) 1.30 (0.7–2.3)
Non-marriage/marriage 260 (6.1) 0.77 (0.5–1.3) 0.64 (0.4–1.1) 0.63 (0.4–1.1) 0.63 (0.4–1.1) 0.63 (0.4–1.1)
Non-marriage/non-marriage 277 (12.3) 1.54 (1.1–2.2) 1.24 (0.8–1.8) 1.25 (0.8–1.9) 1.26 (0.8–1.9) 1.26 (0.8–1.9)
a
The Poisson regressions were made using the sandwich estimator procedure; the outcome was defined as a dummy variable representing two categories: non-obese
(individuals that did not getting obese during the follow-up period—the reference group) and obesity (incidence cases in LNU 2000 BMI 30 kg/m2); exposures, the
social relationships variables, and covariates were included based on LNU 2000. The Swedish Level of Living Survey (9 years of follow-up) ; N(%) ¼ number of observa-
tions and percentage of obesity according to social relationships changes variables; SRI ¼ Social Relationships Index RR ¼ Relative Risk; CI ¼ confidence Interval; Model
1: social relationships changes þ marital status changes þ age þ social class þ BMI in 1991; Model 2: Model 1 þ physical activity ; Model 3: Model 2 þ alcohol con-
sumption ; Model 4: Model 3 þ smoking.
Furthermore, the frequency of visiting friends had an effect on obe- pattern could be the mutual social support and social control that
sity status for both sexes, although in opposite directions. We found married couples usually experience, while divorced and widowed
a positive association between the frequency of visiting friends and individuals lose this type of support, which may lead to an unheal-
being obese among men, whereas the association was reversed thy lifestyle (31). Furthermore, gender differences in how changes
among women. These results could be explained by social control in marital status affects obesity sheds some light on this issue. Only
theory, which states that the lack of a social network influences women who became non-married (divorced or widowed) during the
health behaviors through low social control over deviant health follow-up period had a higher risk of being non-obese compared
behaviors, such as low levels of physical activity and poor nutrition with those who were married at both measurement points. This
(30). However, the factors for or against good health behaviors result aligns with findings from another study (32) that showed a
depend on the characteristics of the social network. Given the sex significant association between becoming non-married and a BMI
differences in our results, we postulate two possible explanations. decrease in a 2-year follow-up period among women. Although the
First, social contact with friends might be qualitatively different for mechanisms linking marriage and obesity are not fully understood,
women and men in relation to how they influence health behavior. some studies have suggested that some relationships are more im-
Women’s social networks may not exclusively support good health portant than others (33,34).
behaviors to the same extent as men’s networks. Second, even
women with poor social contacts are affected by other forms of According to the SRI that was used to summarize changes in social
social control—namely the media—which enforce the high social relationships, the models confirmed gender differences. Among men,
value of maintaining a slim body. levels of social relationships influenced the development of obesity.
The persistence of poor levels of social relationships across the sur-
Marital status was associated with obesity among both men and vey waves had an even more detrimental impact on health. Our
women. Those who reported that they had been divorced or wid- results showed that primarily men who had poor social relationships
owed had a higher risk of developing obesity compared to those during the whole follow-up period (1991-2000) increased their like-
who reported that they remained married. An explanation for this lihood of becoming obese. Contrarily, no effect was detected among
TABLE 5 Regression coefficients and standard errors estimated from multiple linear regressions for the association between
social relationships changes and BMI in LNUa
Social relationships changes Model 1 b (SE) Model 2 b (SE) Model 3 b (SE) Model 4 b (SE) Model 5 b (SE)
Women (n ¼ 1654)
SRI
Level 4 – – – – –
Level 3 0.37 (0.28) 0.34 (0.28) 0.24 (0.28) 0.25 (0.28) 0.23 (0.29)
Level 2 0.34 (0.36) 0.32 (0.36) 0.34 (0.36) 0.34 (0.36) 0.32 (0.36)
Level 1 0.20 (0.57) 0.16 (0.57) 0.07 (0.58) 0.03 (0.56) 0.06 (0.56)
Marital status changes
Marriage/marriage – – – – –
Marriage/non-marriage 0.44 (0.20)b 0.46 (0.20)b 0.53 (0.20)b 0.55 (0.20)b 0.54(0.20)b
Non-marriage/marriage 0.45 (0.20)b 0.45 (0.21)b 0.19 (0.19) 0.18 (0.19) 0.20(0.19)
Non-marriage/non- marriage 0.11 (0.16) 0.12 (0.16) 0.21 (0.16) 0.23 (0.16) 0.22(0.16)
Men (n ¼ 1686)
SRI
Level 4 – – – – –
Level 3 0.38 (0.21)c 0.37 (0.21)c 0.38 (0.21) 0.38 (0.21)c 0.36 (0.21)c
Level 2 0.41 (0.23)c 0.42 (0.23)c 0.43 (0.23) 0.43 (0.23)c 0.41 (0.23)c
Level 1 0.22 (0.31) 0.13 (0.14) 0.23 (0.32) 0.23 (0.32) 0.22 (0.32)
Marital status changes
Marriage/marriage – – – – –
Marriage/non-marriage 0.22 (0.18) 0.23 (0.18) 0.21 (0.18) 0.21 (0.18) 0.17 (0.18)
Non-marriage/marriage 0.24 (0.15) 0.23 (0.15) 0.24 (0.14)c 0.24 (0.14)c 0.22 (0.15)
Non-marriage/non-marriage 0.10 (0.13) 0.13 (0.13) 0.14 (0.13) 0.14 (0.13) 0.19 (0.13)
The linear regressions were made using the BMI in LNU 1991 as an explanatory variable and the continuous explanatory variable (age) center led in the mean; the out-
come was the BMI (kg/m2) in LNU 2000. The covariates were included based on informations provided by LNU 2000. The Swedish Level of Living Survey (9 years of fol-
low-up)
SRI ¼ Social Relationships Index
a
p < 0.001;
b
p < 0.05;
c
p < 0.10 n ¼ number of observations; b ¼ coefficient estimated; SE ¼ estimation error.Model 1: BMI in 1991þsocial relationships index þ marital status changes þ age;
Model 2: Model 1 þ social class; Model 3: Model 2 þ leisure-time physical activity ; Model 4: Model 3 þ alcohol consumption; Model 5: Model 4 þ smoke status.
women, although they are generally more likely to seek and receive to scrutinizing how different dimensions of social relationships con-
emotional support from their social networks in comparison to men. tribute to obesity. Second, our measure on alcohol consumption was
Women also have a larger dependence on social networks (5). Ini- based on a dichotomous question (i.e., do you ever drink beer, wine,
tially, these results may seem contradictory. However, we draw liquor?). Consequently, the variable was very crude and did not con-
upon alternative explanations for why social relationships contribute sider the quantity or frequency of alcohol consumption. Using a
to lower obesity rates among women (35). One of the key themes more specific measure on alcohol consumption could have a differ-
underlying this assumption may be the differential standards of body ent impact on the associations studied but, unfortunately, a variable
image for women versus men. Various authors, working within both with more detailed information on alcohol consumption at both mea-
feminist and broader sociocultural paradigms, have suggested that surement points (1991 and 2000) was not available. Third, the base-
women in Western societies are constantly faced with sociocultural line information is based on data from 1991, and consequently, new
ideals of slimness that are spread through social networks. Such forms of social connectivity could not be considered (e.g., interac-
pressures might lead to dissatisfaction with the body, dietary tion on the Internet, through social media, etc.). In this respect, gen-
restraints (36), and eating disorders (37) that inadvertently lead to eralization of our findings to the present day may be limited due to
higher risks of obesity. changes in the character of social interactions during the last 10-15
years. Fourth, our data analysis was based on only two time points
Our results need to be interpreted in the context of six potential across a 9-year period. Therefore, we have no additional information
methodological limitations. The first involves the use of five differ- between the two time points, and consequently, possible changes
ent indicators of social relationships. These are based on single between the two measurements could not be investigated. Reverse
questions with limited response alternatives, instead of the use of a causality is possible such that increased body weight influences the
validated instrument which could take several dimensions of social social relationship dimensions rather than vice versa. Fifth, BMI
relationships into account at the same time. However, this approach was based on self-reports of weight and height rather than externally
was helpful in disentangling the various potential mechanisms and administrated measurements of weight and height, which are more
reliable. It is possible that individuals underestimated their weight 9. Tomaka J, Thompson S, Palacios R. The relation of social isolation, loneliness, and
social support to disease outcomes among the elderly. J Aging Health 2006;18:
and overestimated their height. Our use of an algorithm, however, 359-384.
may have corrected this information bias and minimized this prob- 10. Strickland OL, Giger JN, Nelson MA, Davis CM. The relationships among stress,
lem. Finally, we had no information concerning body image and, coping, social support, and weight class in premenopausal African American women
at risk for coronary heart disease. J Cardiovasc Nurs 2007;22:272-278.
consequently, could not test whether the ‘‘body image hypothesis’’
11. Christakis NA, Fowler JH. The spread of obesity in a large social network over 32
explained some of the patterns in our findings. years. N Engl J Med 2007;357:370-379.
12. Cohen DA, Finch BK, Bower A, Sastry N. Collective efficacy and obesity: the
In sum, we used longitudinal data from a panel study of the Swedish potential influence of social factors on health. Soc Sci Med 2006;62:769-778.
13. Gebhardt WA, Maes S. Integrating social-psychological frameworks for health
adult population to ascertain the associations between social rela- behavior research. Am J Health Behav 2001;25:528-536.
tionships and obesity among women and men. Our results showed 14. Weinstein SE, Shide DJ, Rolls BJ. Changes in food intake in response to stress in
that various dimensions of social relationships influenced obesity men and women: psychological factors. Appetite 1997;28:7-18.
status during a 9-year period and that these associations were mod- 15. House JS, Landis KR, Umberson D. Social relationships and health. Science 1988;
241:540-545.
erated by the sex of the respondents. Moreover, based on our find- 16. Koropeckyj-Cox T, Pienta AM, Brown TH. Women of the 1950s and the ‘‘norma-
ings, we also discussed some possible mechanisms that may play a tive’’ life course: the implications of childlessness, fertility timing, and marital sta-
role in the links between social relationships and obesity, although tus for psychological well-being in late midlife. Int J Aging Hum Dev 2007;64:
299-330.
we were unable to empirically examine the contribution of these 17. Berkman L, Glass T. Social integration, social networks, social support, and health.
possible mechanisms. In future studies, investigators should follow In: Berkman L, Kawachi I, ed. Social Epidemiology. New York: Oxford University
individuals regularly with repeated measures of both BMI and social Press; 2000.
18. Berkman LF, Syme SL. Social networks, host resistance, and mortality: a nine-year
relationships dimensions (e.g., structural and functional dimensions) follow-up study of Alameda County residents. Am J Epidemiol 1979;109:186-204.
to ascertain the independent effects of these social aspects on obe- 19. Cohen S. Social relationships and health. Am Psychol 2004;59:676-684.
sity. In addition, future studies should also more thoroughly study 20. Colbert AM, Kim KH, Sereika SM, Erlen JA. An examination of the relationships
the mechanisms linking social relationships and the development of among gender, health status, social support, and HIV-related stigma. J Assoc Nurses
AIDS Care 2010;21:302-313.
obesity and whether the operation of these mechanisms differs 21. Umberson D. Gender, marital status and the social control of health behavior. Soc
between men and women.O Sci Med 1992;34:907-917.
22. Kiecolt-Glaser JK, Newton TL. Marriage and health: his and hers. Psychol Bull
2001;127:472-503.
23. Nyholm M, Gullberg B, Merlo J, Lundqvist-Persson C, Rastam L, Lindblad U. The
Acknowledgments validity of obesity based on self-reported weight and height: Implications for popu-
lation studies. Obesity (Silver Spring) 2007;15:197-208.
The authors thank the government agencies from Brazil and Swe- 24. Garn SM, Leonard WR, Hawthorne VM. Three limitations of the body mass index.
Am J Clin Nutr 1986;44:996-997.
den. This study was supported by CAPES—a Brazilian federal gov-
25. Bygren M. Reliabiliteten i Levnadsnivåunders€ okningen - 1991. Stockholm: Stock-
ernment agency research—, FAPERJ—a government agency holm University; 1995.
research from the State of Rio de Janeiro—and The Swedish Foun- 26. Statistics Sweden. Yrkesklassificeringar i FoB 85 enligt Nordisk Yrkesklassificering
dation for International Cooperation in Research and Higher Educa- (NYK) och Socioekonomisk indelning (SEI). MiS 1989:5. (Occupations in Popula-
tion and Housing Census 1985 According to Nordic Standard Occupational Classi-
tion (STINT). fication (NYK) and Swedish Socio-economic Classification (SEI).). Stockholm:
Statistics Sweden; 1989.
C 2013 The Obesity Society
V 27. Warton DI. Regularized sandwich estimators for analysis of high-dimensional data
using generalized estimating equations. Biometrics 2011;67:116-123.
28. Hawkley LC, Cacioppo JT. Loneliness matters: a theoretical and empirical review
References of consequences and mechanisms. Ann Behav Med 2010;40:218-227.
29. Pyszczynski T, Greenberg J, Solomon S, Arndt J, Schimel J. Why do people need
1. Sundquist J, Johansson SE, Sundquist K. Levelling off of prevalence of obesity in
self-esteem? A theoretical and empirical review. Psychol Bull 2004;130:435-468.
the adult population of Sweden between 2000/01 and 2004/05. BMC Public Health
2010;10:119. 30. Feunekes GI, de Graaf C, Meyboom S, van Staveren WA. Food choice and fat
intake of adolescents and adults: associations of intakes within social networks.
2. Oda E, Kawai R. Age- and gender-related differences in correlations between ab-
Prev Med 1998;27:645-656.
dominal obesity and obesity-related metabolic risk factors in Japanese. Intern Med
2009;48:497-502. 31. Barry D, Petry N. Gender differences in associations between stressful life events
and body mass index. Prev Med 2008;47:498-503.
3. Eisenberg D, Quinn BC. Estimating the effect of smoking cessation on weight gain:
32. Jeffery RW, Rick AM. Cross-sectional and longitudinal associations between body
an instrumental variable approach. Health Serv Res 2006;41:2255-2266.
mass index and marriage-related factors. Obes Res 2002;10:809-815.
4. Needham BL, Epel ES, Adler NE, Kiefe C. Trajectories of change in obesity and 33. Sobal J, Nelson MK. Commensal eating patterns: a community study. Appetite
symptoms of depression: the CARDIA study. Am J Public Health 2010;100: 2003;41:181-190.
1040-1046.
34. Tzotzas T, Vlahavas G, Papadopoulou SK, Kapantais E, Kaklamanou D, Hassapi-
5. Borgatti SP, Mehra A, Brass DJ, Labianca G. Network analysis in the social scien- dou M. Marital status and educational level associated to obesity in Greek adults:
ces. Science 2009;323:892-895. data from the National Epidemiological Survey. BMC Public Health 2010;10:732.
6. Friedmann E, Thomas SA, Liu F, Morton PG, Chapa D, Gottlieb SS. Relationship 35. Austin SB, Haines J, Veugelers PJ. Body satisfaction and body weight: gender dif-
of depression, anxiety, and social isolation to chronic heart failure outpatient mor- ferences and sociodemographic determinants. BMC Public Health 2009;9:313.
tality. Am Heart J 2006;152:940 e1-940 e8. 36. Fett AK, Lattimore P, Roefs A, Geschwind N, Jansen A. Food cue exposure and
7. Holt-Lunstad J, Smith TB, Layton JB. Social relationships and mortality risk: a body image satisfaction: the moderating role of BMI and dietary restraint. Body
meta-analytic review. PLoS Med 2010;7:e1000316. Image 2009;6:14-18.
8. Glymour MM, Weuve J, Fay ME, Glass T, Berkman LF. Social ties and cognitive 37. Cobelo AW, de Chermont Prochnik Estima C, Nakano EY, Conti MA, Cordas TA.
recovery after stroke: does social integration promote cognitive resilience? Neuroe- Body image dissatisfaction and eating symptoms in mothers of adolescents with eat-
pidemiology 2008;31:10-20. ing disorders. Eat Weight Disord 2010;15:e219-e225.