Skip to main content Accessibility help
×
Home
Hostname: page-component-5d6d958fb5-zkswk Total loading time: 0.982 Render date: 2022-11-28T18:39:26.886Z Has data issue: true Feature Flags: { "shouldUseShareProductTool": true, "shouldUseHypothesis": true, "isUnsiloEnabled": true, "useRatesEcommerce": false, "displayNetworkTab": true, "displayNetworkMapGraph": false, "useSa": true } hasContentIssue true

Association of Adult Socioeconomic Status with Body Mass Index: A Within- and Between-Twin Study

Published online by Cambridge University Press:  14 April 2021

Samuel J. C. Crofts*
Affiliation:
Twins Research Australia, Centre for Epidemiology and Biostatistics, Melbourne School of Population and Global Health, The University of Melbourne, Melbourne, Victoria, Australia
Janine Lam
Affiliation:
Neuroscience Research Australia, Sydney, New South Wales, Australia School of Psychology, University of New South Wales, Sydney, New South Wales, Australia
Katrina J. Scurrah
Affiliation:
Twins Research Australia, Centre for Epidemiology and Biostatistics, Melbourne School of Population and Global Health, The University of Melbourne, Melbourne, Victoria, Australia Centre for Epidemiology and Biostatistics, The University of Melbourne, Melbourne, Victoria, Australia
Gillian S. Dite
Affiliation:
Genetic Technologies Ltd., Melbourne, Victoria, Australia Centre for Epidemiology and Biostatistics, The University of Melbourne, Melbourne, Victoria, Australia
*
Author for correspondence: Samuel Crofts, Email: sam.crofts@unimelb.edu.au

Abstract

Adult socioeconomic status (SES) has been consistently associated with body mass index (BMI), but it is unclear whether it is linked to BMI independently of childhood SES or other potentially confounding factors. Twin studies can address this issue by implicitly controlling for childhood SES and unmeasured confounders. This co-twin control study used cross-sectional data from Twins Research Australia’s Health and Lifestyle Questionnaire (N = 1918 twin pairs). We investigated whether adult SES, as measured by both the Index of Relative Socioeconomic Disadvantage (IRSD) and the Australian Socioeconomic Index 2006 (AUSEI06), was associated with BMI after controlling for factors shared by twins within a pair. The primary analysis was a linear mixed-effects model that estimated effects both within and between pairs. Between pairs, a 10-unit increase in AUSEI06 was associated with a 0.29 kg/m2 decrease in BMI (95% CI [−.42, −.17], p < .001), and a 1-decile increase in IRSD was associated with a 0.26 kg/m2 decrease in BMI (95% CI [−.35, −.17], p < .001). No association was observed within pairs. In conclusion, higher adult SES was associated with lower BMI between pairs, but no association was observed within pairs. Thus, the link between adult SES and BMI may be due to confounding factors common to twins within a pair.

Type
Articles
Copyright
© The Author(s) 2021

Overweight and obesity are widespread conditions that are commonly measured by thresholds on the body mass index (BMI) (World Health Organization [WHO], 2000). They are major risk factors for cardiovascular disease, diabetes, musculoskeletal disorders and various cancers (WHO, 2010). As such, understanding the causes of overweight and obesity is critical to improving population health, especially given their rapidly increasing prevalence in many countries (Hruby & Hu, Reference Hruby and Hu2015).

It is well documented that socioeconomic status (SES) is associated with overweight and obesity (Ball & Crawford, Reference Ball and Crawford2005; Galobardes et al., Reference Galobardes, Smith and Lynch2006; Herzog et al., Reference Herzog, Lacruz, Haerting, Hartwig, Tiller, Medenwald, Vogt, Thorand, Holle, Bachlechner, Boeing, Merz, Nöthlings, Schlesinger, Schipf, Ittermann, Aumann, Schienkiewitz, Haftenberger and Kluttig2016; Monteiro et al., Reference Monteiro, Moura, Conde and Popkin2004; Tiikkaja et al., Reference Tiikkaja, Olsson, Malki, Modin and Sparén2012; Wu et al., Reference Wu, Ding, Wu, Li, Hu, Hou and Mao2015). Measures of SES commonly reflect a combination of income, education and occupation (Winkleby et al., Reference Winkleby, Jatulis, Frank and Fortmann1992). As a country’s gross national product increases, overweight and obesity move from diseases associated with people of higher SES to diseases associated with people of lower SES (Monteiro et al., Reference Monteiro, Moura, Conde and Popkin2004). In high-income countries, low SES has been consistently associated with an increased risk of childhood overweight and obesity (Wu et al., Reference Wu, Ding, Wu, Li, Hu, Hou and Mao2015). Systematic reviews have also shown that adults of lower SES generally experience greater weight gain over time than adults of higher SES (Ball & Crawford, Reference Ball and Crawford2005; Herzog et al., Reference Herzog, Lacruz, Haerting, Hartwig, Tiller, Medenwald, Vogt, Thorand, Holle, Bachlechner, Boeing, Merz, Nöthlings, Schlesinger, Schipf, Ittermann, Aumann, Schienkiewitz, Haftenberger and Kluttig2016). Finally, there is strong evidence that gender may modify the relationship between SES and BMI (McLaren, Reference McLaren2007; Sobal & Stunkard, Reference Sobal and Stunkard1989). For example, the association between higher SES and lower BMI in high-income countries has generally been more consistent for women than for men (McLaren, Reference McLaren2007; Sobal & Stunkard, Reference Sobal and Stunkard1989). This finding has been attributed to different social pressures and desires between men and women regarding body size (McLaren, Reference McLaren2007; Sobal & Stunkard, Reference Sobal and Stunkard1989).

There are many potential pathways by which SES may be influencing BMI, although there is debate as to which are most valid (Claassen et al., Reference Claassen, Klein, Bratanova, Claes and Corneille2018). The higher cost of healthier diets in some countries and the lack of availability of healthy food choices in low-income areas (along with an abundance of unhealthy food choices) are two prominent theories (Aggarwal et al., Reference Aggarwal, Monsivais, Cook and Drewnowski2011; Story et al., Reference Story, Kaphingst, Robinson-O’Brien and Glanz2008). However, other factors associated with lower SES may also help explain the relationship with BMI. For example, higher stress levels (Moore & Cunningham, Reference Moore and Cunningham2012; Spinosa et al., Reference Spinosa, Christiansen, Dickson, Lorenzetti and Hardman2019) and other psychological factors (Cheon & Hong, Reference Cheon and Hong2017) have been associated with both unhealthy eating patterns and low SES.

Although SES has been consistently associated with BMI in general, the results are less clear when childhood and adult SES are considered separately. In particular, it is not clear whether adult SES is linked to BMI independently of childhood SES. Childhood SES is a predictor of adult SES (d’Addio, Reference d’Addio2007), and disentangling their separate effects is challenging. Additionally, there may be confounding factors, such as genetics, which are responsible for the observed associations between adult SES and BMI (Dinescu et al., Reference Dinescu, Horn, Duncan and Turkheimer2016).

Studies of families and twins are one way of addressing these difficulties. Siblings generally grow up together in the same household and have the same parental figures. Therefore, siblings will usually have the same SES as each other throughout their childhoods. If they leave home and pursue different careers or live in different areas, their SES may change. By comparing siblings with each other, it is possible to isolate the effect of adult SES on BMI while holding childhood SES constant. Importantly, these study designs also allow for the control of other unmeasured factors that are similar between siblings, such as genetics and family environment.

Twin and family studies on similar topics have so far been limited and have given somewhat mixed results. In an Australian study, Webbink et al. (Reference Webbink, Martin and Visscher2010) found that more years of education were linked to lower bodyweight within male twin pairs, but no association was observed within female twin pairs. In another Australian study, Scurrah et al. (Reference Scurrah, Kavanagh, Bentley, Thornton and Harrap2015) found that a change to lower SES in young adulthood was associated with an increase in BMI after controlling for childhood SES. In studies from other countries, Tiikkaja et al. (Reference Tiikkaja, Olsson, Malki, Modin and Sparén2012) found a similar result, in that moving from a higher to lower occupational class was associated with an increase in mortality from cardiovascular disease (for which BMI is a prominent risk factor). Krieger et al. (Reference Krieger, Chen, Coull and Selby2005) also found that lower adult social class in twins was associated with higher cardiovascular disease risk and other measures of poor health, but not with BMI specifically. Interestingly, these relationships did not hold when years of education were used as a measure of adult SES, although others have found a weak association (Fujiwara & Kawachi, Reference Fujiwara and Kawachi2009). Dinescu et al. (Reference Dinescu, Horn, Duncan and Turkheimer2016) found that while adult SES was associated with BMI in their population as a whole, the effect disappeared when looking at the family level. Similarly, Osler et al. (Reference Osler, McGue and Christensen2007) found no significant differences in adult BMI in twins discordant for adult SES.

Our study aims to clarify these varied results by using a twin study design that controls for measured and unmeasured confounders to examine the association between adult SES and BMI. We use a statistical approach in which we include both within- and between-pair effects in the same model. To the best of our knowledge, this method has not been used before in this context and it allows inferences to be made about the potential impact of confounding on the association, which has not been a focus of other published articles on this topic. Additionally, we use both individual and area-level measures of SES, which to the best of our knowledge has only been done once before on a similar topic with a twin design (Dinescu et al., Reference Dinescu, Horn, Duncan and Turkheimer2016). Overweight and obesity are generally preventable and reversible conditions (WHO, 2010), and by further understanding their main drivers, there is significant potential to dramatically reduce their impact.

Materials and Methods

Study Design, Setting and Participants

Cross-sectional data were obtained from Twins Research Australia (TRA), which maintains a registry of twins who are willing to consider participation in ethically approved research (Murphy et al., Reference Murphy, Lam, Cutler, Tyler, Calais-Ferreira, Li, Little, Ferreira, Craig, Scurrah and Hopper2019). Eligible twins were those who had completed the adult version of the TRA Health and Lifestyle Questionnaire, which was emailed to all Australian twin pairs on the registry who were aged 18 years and over and had provided an email address (Cutler, Reference Cutler2015). The questionnaire covered a variety of topics such as health, education, employment and zygosity (Cutler, Reference Cutler2015). All data were self-reported.

Exposure Measures

This study used two indicators of SES: the Australian Socioeconomic Index 2006 (AUSEI06; McMillan, Beavis et al., Reference McMillan, Beavis and Jones2009) and the Index of Relative Socioeconomic Disadvantage (IRSD; Pink, Reference Pink2011).

The AUSEI06 is a socioeconomic index based on occupation data (McMillan, Beavis et al., Reference McMillan, Beavis and Jones2009; Mcmillan, Jones et al., Reference McMillan, Jones and Beavis2009), and it (and its previous iterations) have been widely used as a measure of SES (e.g. Byrne et al. (Reference Byrne, Schwartz, Simmons, Sheeber, Whittle and Allen2018); Lam et al. (Reference Lam, Tyler, Scurrah, Reavley and Dite2019); Nabipour et al., Reference Nabipour, Cumming, Handelsman, Litchfield, Naganathan, Waite, Creasey, Janu, Le Couteur, Sambrook and Seibel2011). It uses occupation data coded according to the Australian and New Zealand Standard Classification of Occupations (ANZSCO) to create scores on a status scale from 0 (low-status occupations) to 100 (high-status occupations) (McMillan, Beavis et al., Reference McMillan, Beavis and Jones2009); see Table 1. We coded responses from the Health and Lifestyle Questionnaire into these categories (Lam et al., Reference Lam, Tyler, Scurrah, Reavley and Dite2019). In all regression analyses, AUSEI06 was used in steps of 10 units for an appropriate effect size.

Table 1. Australian Socioeconomic Index 2006 (AUSEI06) scores for Australian and New Zealand Standard Classification of Occupations (ANZSCO) major groups

Note: *Calculated as the mean score of component units, and it is weighted according to the number of workers in each unit.

Data are sourced from McMillan et al. (Reference McMillan, Beavis and Jones2009).

The IRSD is a more composite, but less specific measure of SES based on area-level variables relating to education, occupation and economic resources (Pink, Reference Pink2011). The IRSD ranks areas from most disadvantaged to least disadvantaged (Pink, Reference Pink2011). IRSD scores have been validated as generally reflecting the SES of regions as determined by the Australian Bureau of Statistics (Pink, Reference Pink2011) and have been associated with BMI in a similar family study (Scurrah et al., Reference Scurrah, Kavanagh, Bentley, Thornton and Harrap2015). Investigating both individual and area-based measures is important because they reflect different aspects of SES and may have different associations with BMI (Chen & Paterson, Reference Chen and Paterson2006). As recommended by the Australian Bureau of Statistics, our analysis used IRSD scores in deciles for ease of interpretation (Pink, Reference Pink2011).

Statistical Analysis

Methods to account for clustering are necessary in twin analyses because outcomes within a pair tend to be correlated. We used a regression model outlined by Carlin et al. (Reference Carlin, Gurrin, Sterne, Morley and Dwyer2005) that accounts for clustering and includes estimates of the association between SES and BMI both within and between twin pairs. This model is known as a co-twin control within-between analysis and can be described as follows:

$${{\bi Y_{ij}}} = {\beta _0} + {\beta _w}({X_{ij}} - \;{\mathop \overline X}_{i}) + \;{\beta _b}{\mathop \overline X}_{i}\comma $$

where $${Y_{ij}}$$ is the expected value of the outcome for a certain individual in the $$i$$ th twin pair, in which $$j$$ equals either 1 or 2 and represents the arbitrary ordering of twins in the pair. β b is the between-pair coefficient, representing the expected change in $$Y$$ for a unit change in the mean value of variable $$X$$ in the $$i$$ th twin pair ( $${\mathop \overline X _i}$$ ). β w is the within-pair coefficient, representing the expected change in $$Y$$ for a unit change in the difference between a twin’s individual value of the exposure variable $$X$$ ( $${X_{ij}}$$ ) and the mean value of variable $$X$$ in their twin pair ( $${\mathop \overline X _i}$$ ). β0 is the intercept.

The within-pair coefficient (β w ) will not be affected by any potentially confounding factors that are common to twins within a pair (Carlin et al., Reference Carlin, Gurrin, Sterne, Morley and Dwyer2005). The between-pair coefficient (β b ) may be affected by these potentially confounding factors, and so a comparison of these two coefficients yields important information on the impact of confounding when not accounting for these factors (Carlin et al., Reference Carlin, Gurrin, Sterne, Morley and Dwyer2005).

This model was fitted using a mixed-effects regression approach, accounting for clustering by incorporating a pair-specific random effect term that allows the intercept to vary between pairs (not shown in equation for simplicity). Age and gender both affect BMI and were adjusted for in all analyses. Parents’ post-school education was also included in the model because it has been commonly used as a proxy measure of childhood SES (Wu et al., Reference Wu, Ding, Wu, Li, Hu, Hou and Mao2015) and was the best measure of this available from the Health and Lifestyle Questionnaire data. Childhood SES is correlated with adult SES and BMI, and hence is a potential confounder in the between-pair analyses (it is controlled for implicitly in the within-pair analyses).

Gender was tested as an effect modifier via likelihood ratio tests. These compared the models with and without an interaction between SES and gender, for both within- and between-pair effects and for both measures of SES.

A sensitivity analysis was conducted to determine the effect of outlying data. Standardized residuals were calculated after fitting the model and any observation with an absolute value greater than 3 was deemed a likely outlier and removed from the analysis (Field, Reference Field2013). The model was then fitted again.

Results

Figure 1 shows the participant flowchart; 1918 twin pairs completed the Health and Lifestyle Questionnaire from approximately 35,000 adult pairs on the registry. After removal of those with incorrect or missing data, 1431 pairs remained for the AUSEI06 analysis and 1558 remained for the IRSD analysis.

Fig. 1. Participant flowchart.

Table 2 shows the characteristics of the study participants. As the continuous variables were somewhat skewed, their distributions are displayed as median (interquartile range) rather than mean (standard deviation). Participants were predominantly female (76.6%) and monozygotic (74.1%). Ages ranged from 18 to 85, with a median of 48 (interquartile range 35–60). The median BMI was 24.5 kg/m2, with just under half of all participants classified as overweight or obese by the WHO thresholds (WHO, 2000). This is slightly below the national average of 63% (Australian Institute of Health and Welfare, 2018). Average SES was higher than in the general population, with a median IRSD decile of 8 and median AUSEI06 score of 67 — a score that lies between the two highest ANZSCO groups of managers (58.1) and professionals (81.6) (McMillan, Beavis et al., Reference McMillan, Beavis and Jones2009); see Table 1.

Table 2. Characteristics of participants in the sample after removal of those with missing or incorrect outcome data (N = 3526 individuals)

There was little correlation between age and SES in the sample (r = .05 between age and AUSEI06 and r = −.05 between age and IRSD decile). There was some difference in mean SES between genders: mean (SD) AUSEI06 score 62.4 (21.8) for females and 64.1 (22.4) for males, and mean (SD) IRSD decile 6.7 (2.7) for females and 7.1 (2.6) for males. Similarly, mean SES varied depending on whether participants’ parents had undertaken post-school education: mean (SD) AUSEI06 score 60.9 (21.7) for neither parent having undertaken post-school education and 66.5 (21.6) for at least one parent, mean (SD) IRSD decile 6.4 (2.8) for neither parent and 7.1 (2.6) for at least one parent.

Approximately 65% of twin pairs were discordant for IRSD deciles and the median within-pair absolute difference in the AUSEI06 score was 11.9 (interquartile range 3.0–28.7). The median within-pair difference in BMI was 2.0 kg/m2 (interquartile range 0.8–4.1).

Figures 2 and 3 give a visual representation of the range of within-pair differences in the exposures and the outcome, and the unadjusted relationship between them. As shown by the flat overlaid Lowess lines, no association was observed between within-pair differences in either measure of SES and BMI.

Fig. 2. Scatter plot of within-pair differences in body mass index versus within-pair differences in Australian Socioeconomic Index 2006 score with Lowess line overlaid.

Fig. 3. Scatter plot of within-pair differences in body mass index versus within-pair differences in Index of Relative Socioeconomic Disadvantage decile with Lowess line overlaid.

Table 3 shows the results of the co-twin control within-between analysis, which adjusted for age, gender and parents’ post-school education. Consistent with Figures 2 and 3, the within-pair estimates (β w ) for both SES measures were approximately zero and had confidence intervals roughly centered around the null, showing that no association between SES and BMI was observed within-pairs. Conversely, the between-pair estimates (β b ) showed strong inverse associations between SES and BMI. A 10-unit increase AUSEI06 was associated with an estimated population mean BMI decrease of 0.29 kg/m2 (95% CI [−.42, −.17], p < .001) and a 1-decile increase in IRSD score was associated with a population mean BMI decrease of 0.26 kg/m2 (95% CI [−.35, −.17], p < .001).

Table 3. Results of the co-twin control within-between analysis of the effect of two measures of socioeconomic status on body mass index (kg/m2)

β w = within-pair coefficient; β b = between-pair coefficient.

Note: Model fit using a mixed-effects approach and is adjusted for age, gender and parents’ post-school education.

The likelihood ratio tests comparing the models with and without a gender interaction provided no evidence against the null hypothesis of no modification of the relationship between SES and BMI by gender (p = .81 for the AUSEI06 model, p = .30 for the IRSD model).

No meaningful differences were observed in the results after removing potential outliers in the sensitivity analysis (data are not shown). Additionally, as there was such an extreme imbalance in the zygosity and gender of pairs, all analyses were repeated with just monozygotic pairs and separately with just female–female pairs. No meaningful differences were observed in either case. For example, upon restriction to just female–female pairs, the within-pair and between-pair coefficients for the AUSEI06 regression changed from .01 (95% CI [−.10, .13]) and −.29 (95% CI [−.42, −.17]), respectively, to −.02 (95% CI [−.16, .11]) and −.30 (95% CI [−.45, −.15]). Upon restriction to just monozygotic pairs, these coefficients changed to .06 (95% CI [−.07, .19]) and −.34 (95% CI [−.48, −.19]). As such, we did not restrict our sample to these categories in the main analyses.

Discussion

Main Findings

The between-pair coefficients give estimates that do not account for unmeasured potential confounders that are common to twins within a pair. As shown in Table 3, these values indicate moderately sized, but strong inverse associations between both measures of adult SES and BMI between twin pairs.

In contrast, the within-pair coefficients account for any unmeasured confounders that are common to twins within a pair. As shown in Table 3, the magnitudes of these estimates were all approximately zero and provided no evidence against the null hypotheses of no association. This result shows that there was no association between adult SES and BMI once we implicitly controlled for these unmeasured confounders.

Together, the results from the within- and between-pair analyses suggest that substantial confounding is present in the association between adult SES and BMI when not controlling for factors common to twins within a pair.

Finally, despite previous findings that gender may be an effect modifier of the relationship between SES and BMI (McLaren, Reference McLaren2007; Sobal & Stunkard, Reference Sobal and Stunkard1989), our study does not support this conclusion.

Interpretations

Our finding of no association between adult SES and BMI after accounting for factors shared between twins is one that is consistent with much of the previous research. Of the prominent twin and family studies looking into SES and BMI, Osler et al. (Reference Osler, McGue and Christensen2007) and Dinescu et al. (Reference Dinescu, Horn, Duncan and Turkheimer2016) also found no association. Krieger et al. (Reference Krieger, Chen, Coull and Selby2005) did not find an association between adult SES and BMI either, although they did find an association with cardiovascular disease risk, as did Tiikkaja et al. (Reference Tiikkaja, Olsson, Malki, Modin and Sparén2012). One explanation for this finding may be that adult SES is associated with cardiovascular disease via pathways that do not involve BMI. Somewhat consistent with our findings, Webbink et al. (Reference Webbink, Martin and Visscher2010) found similar null results when looking at years of education and adult BMI within female twin pairs, although they did find an association within male twin pairs. However, our study used more direct measures of SES than years of education, which may explain this difference. Finally, Scurrah et al. (Reference Scurrah, Kavanagh, Bentley, Thornton and Harrap2015) found that a change in SES in young adulthood is inversely associated with BMI while controlling for childhood SES. Despite their study also being conducted on an Australian population, their results are not consistent with our findings. However, some significant differences may help explain this discrepancy, such as the fact that the study occurred approximately 20 years before the release of the Health and Lifestyle Questionnaire (1990−1996 compared to 2014 onwards) and involved a much younger sample (18−30 years compared to 18 years and over).

Our findings suggest that there are factors common to twins within a pair that are confounding associations between adult SES and BMI seen in the general population (i.e. between unrelated individuals). It is unclear from the literature exactly what these factors are, but there are many potential candidates. Early life environment is one possibility and is known to be a strong influence on both adult SES (Branigan et al., Reference Branigan, McCallum and Freese2013) and adult BMI (Campbell, Reference Campbell2016). If any aspects of early life environment affect both variables, then confounding will occur. For example, an authoritative parenting style, characterized by rational, warm and responsive direction (Baumrind, Reference Baumrind1967), has been associated with lower BMI (Sokol et al., Reference Sokol, Qin and Poti2017). It has also been associated with greater academic success (Spera, Reference Spera2005; Steinberg et al., Reference Steinberg, Lamborn, Dornbusch and Darling1992), which is linked to measures of adult SES such as educational attainment. Combined, these two associations would potentially result in the creation of a noncausal association between adult SES and BMI.

Similarly, genes may play a confounding role in the association between adult SES and BMI (Dinescu et al., Reference Dinescu, Horn, Duncan and Turkheimer2016). It is well established that there is a heritable component to BMI (Goodarzi, Reference Goodarzi2018; Locke et al., Reference Locke, Kahali, Berndt, Justice, Pers, Day, Powell, Vedantam, Buchkovich, Yang, Croteau-Chonka, Esko, Fall, Ferreira, Gustafsson, Kutalik, Luan, Mägi, Randall and Speliotes2015) and to certain measures of adult SES, such as educational attainment (Branigan et al., Reference Branigan, McCallum and Freese2013; Lee et al., Reference Lee, Wedow, Okbay, Kong, Maghzian, Zacher, Nguyen-Viet, Bowers, Sidorenko, Karlsson Linnér, Fontana, Kundu, Lee, Li, Li, Royer, Timshel, Walters, Willoughby and Cesarini2018; Marks, Reference Marks2017; von Stumm et al., Reference von Stumm, Smith-Woolley, Ayorech, McMillan, Rimfeld, Dale and Plomin2020). As such, confounding will occur if there is a common genetic component between the two variables. This may work through many possible pathways, including psychological factors. For example, a recent meta-analysis concluded that there is a genetic basis to self-discipline (Willems et al., Reference Willems, Boesen, Li, Finkenauer and Bartels2019). This is a trait that may affect dietary choices and consequently BMI (Koike et al., Reference Koike, Hardy and Richards2016; Tsukayama et al., Reference Tsukayama, Toomey, Faith and Duckworth2010). Additionally, self-discipline may affect academic success (Duckworth & Seligman, Reference Duckworth and Seligman2005; Mischel et al., Reference Mischel, Shoda and Rodriguez1989) and educational attainment (Moffitt et al., Reference Moffitt, Arseneault, Belsky, Dickson, Hancox, Harrington, Houts, Poulton, Roberts, Ross, Sears, Thomson and Caspi2011), and consequently adult SES. Again, such a scenario would create an association between adult SES and BMI in the absence of any causal relationship.

Although our study does not support a causal association between SES and BMI in adulthood, other literature does suggest that one may exist in childhood and adolescence. As mentioned previously, a recent systematic review concluded that children of lower SES backgrounds in high-income countries had significantly higher risks of overweight and obesity (Wu et al., Reference Wu, Ding, Wu, Li, Hu, Hou and Mao2015). Other systematic reviews have found similar associations between childhood SES and adult cardiovascular disease (Galobardes et al., Reference Galobardes, Smith and Lynch2006; Pollitt et al., Reference Pollitt, Rose and Kaufman2005), a relationship in which BMI may be on the causal pathway (Pollitt et al., Reference Pollitt, Rose and Kaufman2005).

In explaining the association between childhood SES and adult cardiovascular disease in their systematic review, Pollitt et al. (Reference Pollitt, Rose and Kaufman2005) found most support for the theory that the psychosocial and physiological impacts associated with low SES can accumulate throughout a person’s life and result in negative health outcomes. Additionally, many of the links between early life SES and BMI may exist through learned behaviors, such as diet and exercise practices (Pollitt et al., Reference Pollitt, Rose and Kaufman2005), which may persist in later life despite a change in SES. Unsurprisingly, childhood obesity is also a very strong risk factor for adulthood obesity (WHO, 2000). Taken together, these findings lead to the conclusion that any change in SES in adulthood will be operating on strong foundations laid in early life and that these may be difficult to change. These early life foundations are perhaps why we did not find an association between adult SES and BMI, but one does seem to exist between childhood SES and BMI.

Strengths and Limitations

Our study had several significant strengths. Most importantly, the twin study design provides the ability to account for confounding from unmeasured variables that are common to twins within a pair. Our statistical approach, in which we included both within- and between-pair effects in the same model, to our knowledge had not been used before in this context and allowed us to make inferences about the potential impact of confounding on the association between adult SES and BMI. Additionally, our use of two separate indicators of SES is valuable given that one indicator may not show the whole picture, and that the relationship with BMI can vary with different measures (Basto-Abreu et al., Reference Basto-Abreu, Barrientos-Gutiérrez, Zepeda-Tello, Camacho, Gimeno Ruiz de Porras and Hernández-Ávila2018; Galobardes et al., Reference Galobardes, Smith and Lynch2006; Krieger et al., Reference Krieger, Chen, Coull and Selby2005; McLaren, Reference McLaren2007). Our conclusions are consistent across both an individual and an area-level measure of SES, which increases their generalizability to SES as a whole. The sample size was also large — especially for a twin study — which gives weight to the results. Finally, the sensitivity analysis conducted also showed that the results are robust and not significantly affected by the presence of any outliers.

The main limitation of this study was that the sample was quite homogenous. Of particular note, there were few people from the lowest SES backgrounds, which may make conclusions regarding them less strong. That said, the sample size was large enough that most underrepresented groups still had a reasonable number of people in them. Similarly, as is often the case with twin studies in general, the differences in exposures within twin pairs tended to be relatively small, which may have increased the chance of a null finding regarding the within-pair estimates. However, the confidence intervals for the within-pair estimates were quite narrow, with the largest magnitude limits being an increase in BMI of 0.13 kg/m2 for every 10 units increase in AUSEI06 and a 0.14 kg/m2 decrease for every decile increase in IRSD. Hence, large true effects appear implausible from these results, despite the relatively small differences in exposures within twin pairs.

Additionally, there may be some selection bias occurring within this sample due to the nature of the questionnaire. As is the case with optional participation, people who respond may be systematically different from those who do not. In this case, it is reasonable to suspect that those who responded may be more likely to have an interest in health research, and by extension, their own health.

Finally, we were somewhat limited by the availability of variables in our dataset. The proxy measure of childhood SES we used (parents’ post-school education) is by no means perfect, and so there may be some residual confounding in the between-twin estimates. Unfortunately, more accurate measures of childhood SES such as parental income were not available. Similarly, a measure of household income, which could have been used in addition to the AUSEI06 as a measure of individual SES, was also unavailable. The fact that the AUSEI06 only applies to part- and full-time workers and reflects only an individual’s status (as opposed to the highest status in the household) are drawbacks of this measure.

Conclusions

This study found that associations observed between adult SES and BMI may be explained by factors that are common to twins within a pair. This has significant implications for research into adult SES and BMI that does not account for potential confounding factors. Future research should determine what these confounding factors are so they can potentially be targeted in health interventions.

Acknowledgments

I would like to extend my thanks to Twins Research Australia for the opportunity to conduct this research, along with all the generous participants of the Health and Lifestyle Questionnaire who made it possible.

Financial Support

This research was facilitated through access to Twins Research Australia, a national resource supported by a Centre of Research Excellence Grant (ID: 1079102), from the National Health and Medical Research Council.

Conflicts of Interest

None.

Ethical Standards

This study received formal approval from TRA and ethics approval from the Melbourne School of Population and Global Health’s Human Ethics Advisory Group (Ethics ID: 1853456.1).

References

Aggarwal, A., Monsivais, P., Cook, A. J., & Drewnowski, A. (2011). Does diet cost mediate the relation between socioeconomic position and diet quality? European Journal of Clinical Nutrition, 65, 10591066.CrossRefGoogle ScholarPubMed
Australian Institute of Health and Welfare. (2018). Australia’s Health 2018. https://www.aihw.gov.au/reports/australias-health/australias-health-2018/contents/table-of-contents Google Scholar
Ball, K., & Crawford, D. (2005). Socioeconomic status and weight change in adults: A review. Social Science & Medicine, 60, 19872010.CrossRefGoogle ScholarPubMed
Basto-Abreu, A., Barrientos-Gutiérrez, T., Zepeda-Tello, R., Camacho, V., Gimeno Ruiz de Porras, D., & Hernández-Ávila, M. (2018). The relationship of socioeconomic status with body mass index depends on the socioeconomic measure used. Obesity, 26, 176184.CrossRefGoogle Scholar
Baumrind, D. (1967). Child care practices anteceding three patterns of preschool behavior. Genetic Psychology Monographs, 75, 4388.Google ScholarPubMed
Branigan, A. R., McCallum, K. J., & Freese, J. (2013). Variation in the heritability of educational attainment: An international meta-analysis. Social Forces, 92, 109140.CrossRefGoogle Scholar
Byrne, M. L., Schwartz, O. S., Simmons, J. G., Sheeber, L., Whittle, S., & Allen, N. B. (2018). Duration of breastfeeding and subsequent adolescent obesity: Effects of maternal behavior and socioeconomic status. Journal of Adolescent Health, 62, 471479.CrossRefGoogle ScholarPubMed
Campbell, M. K. (2016). Biological, environmental, and social influences on childhood obesity. Pediatric Research, 79, 205211.CrossRefGoogle ScholarPubMed
Carlin, J. B., Gurrin, L. C., Sterne, J. A., Morley, R., & Dwyer, T. (2005). Regression models for twin studies: A critical review. International Journal of Epidemiology, 34, 10891099.CrossRefGoogle ScholarPubMed
Chen, E., & Paterson, L. Q. (2006). Neighborhood, family, and subjective socioeconomic status: How do they relate to adolescent health? Health Psychology, 25, 704.CrossRefGoogle ScholarPubMed
Cheon, B. K., & Hong, Y.-Y. (2017). Mere experience of low subjective socioeconomic status stimulates appetite and food intake. Proceedings of the National Academy of Sciences, 114, 7277.CrossRefGoogle Scholar
Claassen, M. A., Klein, O., Bratanova, B., Claes, N., & Corneille, O. (2018). A systematic review of psychosocial explanations for the relationship between socioeconomic status and body mass index. Appetite, 132, 208221.CrossRefGoogle ScholarPubMed
Cutler, T. (2015). The Australian Twin Registry Health and Lifestyle Questionnaire. Australian Twin Registry. http://twins.org.au/images/general/publications/HLQ_abstract.pdf Google Scholar
d’Addio, A. C. (2007). Intergenerational transmission of disadvantage: Mobility or immobility across Generations? A review of the evidence for OECD countries. OECD. https://www.oecd-ilibrary.org/social-issues-migration-health/intergenerational-transmission-of-disadvantage_217730505550 Google Scholar
Dinescu, D., Horn, E. E., Duncan, G., & Turkheimer, E. (2016). Socioeconomic modifiers of genetic and environmental influences on body mass index in adult twins. Health Psychology, 35, 157.CrossRefGoogle ScholarPubMed
Duckworth, A. L., & Seligman, M. E. (2005). Self-discipline outdoes IQ in predicting academic performance of adolescents. Psychological Science, 16, 939944.CrossRefGoogle ScholarPubMed
Field, A. (2013). Discovering statistics using IBM SPSS statistics. SAGE Publications.Google Scholar
Fujiwara, T., & Kawachi, I. (2009). Is education causally related to better health? A twin fixed-effect study in the USA. International Journal of Epidemiology, 38, 13101322.CrossRefGoogle ScholarPubMed
Galobardes, B., Smith, G. D., & Lynch, J. W. (2006). Systematic review of the influence of childhood socioeconomic circumstances on risk for cardiovascular disease in adulthood. Annals of Epidemiology, 16, 91104.CrossRefGoogle ScholarPubMed
Goodarzi, M. O. (2018). Genetics of obesity: What genetic association studies have taught us about the biology of obesity and its complications. Lancet Diabetes Endocrinology, 6, 223236.CrossRefGoogle ScholarPubMed
Herzog, B., Lacruz, M. E., Haerting, J., Hartwig, S., Tiller, D., Medenwald, D., Vogt, S., Thorand, B., Holle, R., Bachlechner, U., Boeing, H., Merz, B., Nöthlings, U., Schlesinger, S., Schipf, S., Ittermann, T., Aumann, N., Schienkiewitz, A., Haftenberger, M., & Kluttig, A. (2016). Socioeconomic status and anthropometric changes — A meta-analytic approach from seven German cohorts. Obesity, 24, 710718.CrossRefGoogle ScholarPubMed
Hruby, A., & Hu, F. B. (2015). The epidemiology of obesity: A big picture. PharmacoEconomics, 33, 673689.CrossRefGoogle ScholarPubMed
Koike, S., Hardy, R., & Richards, M. (2016). Adolescent self-control behavior predicts body weight through the life course: A prospective birth cohort study. International Journal of Obesity, 40, 7176.CrossRefGoogle ScholarPubMed
Krieger, N., Chen, J. T., Coull, B. A., & Selby, J. V. (2005). Lifetime socioeconomic position and twins’ health: An analysis of 308 pairs of United States women twins. PLoS Medicine, 2, e162.CrossRefGoogle ScholarPubMed
Lam, J. R., Tyler, J., Scurrah, K. J., Reavley, N. J., & Dite, G. S. (2019). The association between socioeconomic status and psychological distress: A within and between twin study. Twin Research and Human Genetics, 22, 312320.CrossRefGoogle ScholarPubMed
Lee, J. J., Wedow, R., Okbay, A., Kong, E., Maghzian, O., Zacher, M., Nguyen-Viet, T. A., Bowers, P., Sidorenko, J., Karlsson Linnér, R., Fontana, M. A., Kundu, T., Lee, C., Li, H., Li, R., Royer, R., Timshel, P. N., Walters, R. K., Willoughby, E. A., & Cesarini, D. (2018). Gene discovery and polygenic prediction from a genome-wide association study of educational attainment in 1.1 million individuals. Nature Genetics, 50, 11121121.CrossRefGoogle ScholarPubMed
Locke, A. E., Kahali, B., Berndt, S. I., Justice, A. E., Pers, T. H., Day, F. R., Powell, C., Vedantam, S., Buchkovich, M. L., Yang, J., Croteau-Chonka, D. C., Esko, T., Fall, T., Ferreira, T., Gustafsson, S., Kutalik, Z., Luan, J., Mägi, R., Randall, J. C., & Speliotes, E. K. (2015). Genetic studies of body mass index yield new insights for obesity biology. Nature, 518, 197206.CrossRefGoogle ScholarPubMed
Marks, G. N. (2017). The contribution of genes and the environment to educational and socioeconomic attainments in Australia. Twin Research and Human Genetics, 20, 281289.CrossRefGoogle ScholarPubMed
McLaren, L. (2007). Socioeconomic status and obesity. Epidemiologic Reviews, 29, 2948.CrossRefGoogle ScholarPubMed
McMillan, J., Beavis, A., & Jones, F. L. (2009). The AUSEI06: A new socioeconomic index for Australia. Journal of Sociology, 45, 123149.CrossRefGoogle Scholar
McMillan, J., Jones, F. L., & Beavis, A. (2009). A new scale for measuring socioeconomic status in educational research: Development and validation of the Australian Socioeconomic Index 2006 (AUSEI06). Paper presented at the AARE International Education Research Conference, Canberra.Google Scholar
Mischel, W., Shoda, Y., & Rodriguez, M. I. (1989). Delay of gratification in children. Science, 244, 933938.CrossRefGoogle ScholarPubMed
Moffitt, T. E., Arseneault, L., Belsky, D., Dickson, N., Hancox, R. J., Harrington, H., Houts, R., Poulton, R., Roberts, B. W., Ross, S., Sears, M. R., Thomson, W. M., & Caspi, A. (2011). A gradient of childhood self-control predicts health, wealth, and public safety. Proceedings of the National Academy of Sciences, 108, 26932698.CrossRefGoogle ScholarPubMed
Monteiro, C. A., Moura, E. C., Conde, W. L., & Popkin, B. M. (2004). Socioeconomic status and obesity in adult populations of developing countries: A review. Bulletin of the World Health Organization, 82, 940946.Google ScholarPubMed
Moore, C. J., & Cunningham, S. A. (2012). Social position, psychological stress, and obesity: A systematic review. Journal of the Academy of Nutrition and Dietetics, 112, 518526.CrossRefGoogle ScholarPubMed
Murphy, K., Lam, J., Cutler, T., Tyler, J., Calais-Ferreira, L., Li, S., Little, C., Ferreira, P., Craig, J. M., Scurrah, K. J., & Hopper, J. L. (2019). Twins Research Australia: A new paradigm for driving twin research. Twin Research and Human Genetics, 22, 438445.CrossRefGoogle ScholarPubMed
Nabipour, I., Cumming, R., Handelsman, D. J., Litchfield, M., Naganathan, V., Waite, L., Creasey, H., Janu, M., Le Couteur, D., Sambrook, P. N., & Seibel, M. J. (2011). Socioeconomic status and bone health in community-dwelling older men: The CHAMP study. Osteoporosis International, 22, 13431353.CrossRefGoogle ScholarPubMed
Osler, M., McGue, M., & Christensen, K. (2007). Socioeconomic position and twins’ health: A life-course analysis of 1266 pairs of middle-aged Danish twins. International Journal of Epidemiology, 36, 7783.CrossRefGoogle ScholarPubMed
Pollitt, R. A., Rose, K. M., & Kaufman, J. S. (2005). Evaluating the evidence for models of life course socioeconomic factors and cardiovascular outcomes: A systematic review. BMC Public Health, 5, 7.CrossRefGoogle ScholarPubMed
Scurrah, K. J., Kavanagh, A. M., Bentley, R. J., Thornton, L. E., & Harrap, S. B. (2015). Socioeconomic position in young adulthood is associated with BMI in Australian families. Journal of Public Health, 38, e39e46.CrossRefGoogle ScholarPubMed
Sobal, J., & Stunkard, A. J. (1989). Socioeconomic status and obesity: A review of the literature. Psychological Bulletin, 105, 260.CrossRefGoogle ScholarPubMed
Sokol, R. L., Qin, B., & Poti, J. M. (2017). Parenting styles and body mass index: A systematic review of prospective studies among children. Obesity Reviews, 18, 281292.CrossRefGoogle ScholarPubMed
Spera, C. (2005). A review of the relationship among parenting practices, parenting styles, and adolescent school achievement. Educational Psychology Review, 17, 125146.CrossRefGoogle Scholar
Spinosa, J., Christiansen, P., Dickson, J. M., Lorenzetti, V., & Hardman, C. A. (2019). From socioeconomic disadvantage to obesity: The mediating role of psychological distress and emotional eating. Obesity, 27, 559564.CrossRefGoogle ScholarPubMed
Steinberg, L., Lamborn, S. D., Dornbusch, S. M., & Darling, N. (1992). Impact of parenting practices on adolescent achievement: Authoritative parenting, school involvement, and encosuragement to succeed. Child Development, 63, 12661281.CrossRefGoogle ScholarPubMed
Story, M., Kaphingst, K. M., Robinson-O’Brien, R., & Glanz, K. (2008). Creating healthy food and eating environments: Policy and environmental approaches. Annual Review of Public Health, 29, 253272.CrossRefGoogle ScholarPubMed
Tiikkaja, S., Olsson, M., Malki, N., Modin, B., & Sparén, P. (2012). Familial risk of premature cardiovascular mortality and the impact of intergenerational occupational class mobility. Social Science & Medicine, 75, 18831890.CrossRefGoogle ScholarPubMed
Tsukayama, E., Toomey, S. L., Faith, M. S., & Duckworth, A. L. (2010). Self-control as a protective factor against overweight status in the transition from childhood to adolescence. Archives of Pediatrics & Adolescent Medicine, 164, 631635.CrossRefGoogle ScholarPubMed
von Stumm, S., Smith-Woolley, E., Ayorech, Z., McMillan, A., Rimfeld, K., Dale, P. S., & Plomin, R. (2020). Predicting educational achievement from genomic measures and socioeconomic status. Developmental Science, 23, e12925.Google ScholarPubMed
Webbink, D., Martin, N. G., & Visscher, P. M. (2010). Does education reduce the probability of being overweight? Journal of Health Economics, 29, 2938.CrossRefGoogle ScholarPubMed
Willems, Y., Boesen, N., Li, J., Finkenauer, C., & Bartels, M. (2019). The heritability of self-control: A meta-analysis. Neuroscience & Biobehavioral Reviews, 100, 324334.CrossRefGoogle ScholarPubMed
Winkleby, M. A., Jatulis, D. E., Frank, E., & Fortmann, S. P. (1992). Socioeconomic status and health: How education, income, and occupation contribute to risk factors for cardiovascular disease. American Journal of Public Health, 82, 816820.CrossRefGoogle ScholarPubMed
World Health Organization (WHO). (2000). Obesity: Preventing and managing the global epidemic. https://www.who.int/nutrition/publications/obesity/WHO_TRS_894/en/ Google Scholar
World Health Organization (WHO). (2010). Global status report on noncommunicable diseases 2010. https://www.who.int/nmh/publications/ncd_report2010/en/ Google Scholar
Wu, S., Ding, Y., Wu, F., Li, R., Hu, Y., Hou, J., & Mao, P. (2015). Socio-economic position as an intervention against overweight and obesity in children: A systematic review and meta-analysis. Scientific Reports, 5, 11354.CrossRefGoogle ScholarPubMed
Figure 0

Table 1. Australian Socioeconomic Index 2006 (AUSEI06) scores for Australian and New Zealand Standard Classification of Occupations (ANZSCO) major groups

Figure 1

Fig. 1. Participant flowchart.

Figure 2

Table 2. Characteristics of participants in the sample after removal of those with missing or incorrect outcome data (N = 3526 individuals)

Figure 3

Fig. 2. Scatter plot of within-pair differences in body mass index versus within-pair differences in Australian Socioeconomic Index 2006 score with Lowess line overlaid.

Figure 4

Fig. 3. Scatter plot of within-pair differences in body mass index versus within-pair differences in Index of Relative Socioeconomic Disadvantage decile with Lowess line overlaid.

Figure 5

Table 3. Results of the co-twin control within-between analysis of the effect of two measures of socioeconomic status on body mass index (kg/m2)

You have Access

Save article to Kindle

To save this article to your Kindle, first ensure coreplatform@cambridge.org is added to your Approved Personal Document E-mail List under your Personal Document Settings on the Manage Your Content and Devices page of your Amazon account. Then enter the ‘name’ part of your Kindle email address below. Find out more about saving to your Kindle.

Note you can select to save to either the @free.kindle.com or @kindle.com variations. ‘@free.kindle.com’ emails are free but can only be saved to your device when it is connected to wi-fi. ‘@kindle.com’ emails can be delivered even when you are not connected to wi-fi, but note that service fees apply.

Find out more about the Kindle Personal Document Service.

Association of Adult Socioeconomic Status with Body Mass Index: A Within- and Between-Twin Study
Available formats
×

Save article to Dropbox

To save this article to your Dropbox account, please select one or more formats and confirm that you agree to abide by our usage policies. If this is the first time you used this feature, you will be asked to authorise Cambridge Core to connect with your Dropbox account. Find out more about saving content to Dropbox.

Association of Adult Socioeconomic Status with Body Mass Index: A Within- and Between-Twin Study
Available formats
×

Save article to Google Drive

To save this article to your Google Drive account, please select one or more formats and confirm that you agree to abide by our usage policies. If this is the first time you used this feature, you will be asked to authorise Cambridge Core to connect with your Google Drive account. Find out more about saving content to Google Drive.

Association of Adult Socioeconomic Status with Body Mass Index: A Within- and Between-Twin Study
Available formats
×
×

Reply to: Submit a response

Please enter your response.

Your details

Please enter a valid email address.

Conflicting interests

Do you have any conflicting interests? *