Obesity and its sequelae comprise an

Transcription

1 Opposing socioeconomic gradients in overweight and obese adults Abstract Objective: Investigate the relationship between socioeconomic status (SES) and prevalence of overweight and/or obesity, by sex, using total annual household income as the indicator of SES and the World Health Organization (WHO) recommended ranges of self-reported Body Mass Index (BMI) as the indicator of overweight and/or obesity. Methods: Total annual household income and BMI data were obtained from the Victorian Population Health Survey (VPHS), an annual computer-assisted telephone survey of the health and well-being of Victorian adults aged 18 years and older. Statistical analysis was conducted using ordinary least squares linear regression on the logarithms of age-standardised prevalence estimates of overweight ( kg/m 2 ), obesity ( 30.0 kg/m 2 ), and overweight and obesity combined ( 25.0 kg/m 2 ), by income category and sex. Results: Typical SES gradients were observed in obese males and females, where the prevalence of obesity decreased with increasing income. No SES gradient was observed in overweight females, however, a reverse SES gradient was observed in overweight males, where the prevalence of overweight increased with increasing income. Combining the overweight and obesity categories into a single group eliminated the typical SES gradients observed in males and females for obesity, and resulted in a statistically significant reverse SES gradient in males. Conclusions: Combining the BMI categories of overweight and obesity into a single category masks important SES differences, while combining the data for males and females masks important sex differences. BMI categories of overweight and obesity should be analysed and reported independently, as should BMI data by sex. Key words: Overweight, obesity, gender socioeconomic status; BMI; health inequalities Aust NZ J Public Health. 2013; 37:32-8 doi: / Alison Markwick, Loretta Vaughan Health Intelligence Unit, Prevention and Population Health Branch, Department of Health, Victoria Zahid Ansari Health Intelligence Unit, Prevention and Population Health Branch, Department of Health, Victoria; School of Public Health and Preventive Medicine, Monash University, Victoria. Obesity and its sequelae comprise an increasing and significant health burden in all developed nations. 1 There is a pressing need to reverse this trend, as it is not inconceivable that the gains in life expectancy that we have enjoyed over the last several decades could level off or even reverse as a consequence. 2 Therefore, there is a high-priority need to identify and implement effective interventions in high-risk populations. In Australia, it is common practice to measure obesity using body mass index (BMI), and to categorise a person s body weight status using the World Health Organization (WHO) recommended BMI cutoffs. 1 It is also common practice to combine the two categories of overweight (preobesity) and obesity as an overall indicator of unhealthy weight, and to use this indicator to determine strategic targets for the reduction in prevalence of unhealthy weight. 3,4,5 Most diseases and health conditions tend to follow a typical socioeconomic (SES) gradient, where poorer health outcomes are associated with lower SES. It is relatively rare to find the opposite, or reverse gradient, where the higher a person is on the socioeconomic scale, the worse the health outcomes. Obesity is one such health outcome for which reverse, rather than typical, SES gradients have been observed. 6 However, females in developed countries show strong and consistent typical SES gradients irrespective of the index of SES or obesity used, while the findings for males and children are inconsistent. 7 In this paper, we offer an alternative explanation for the inconsistent findings in males that may lie in the different definitions of obesity employed. Based on the assumption that overweight (or pre-obesity) and obesity are simply on a continuum in terms of risk to health, the two categories are frequently merged and reported as a single statistic, as is commonly practised in Australia. This assumes that the same set of underlying determinants of health apply to both the overweight and obese categories of unhealthy weight; and we will show that this is incorrect. Given that public health policy in Australia is usually mindful of social inequalities in health, and often seeks to address the avoidable inequalities that are due to the unfair distribution of social and economic resources, it is important to have a thorough understanding of any given health indicator and its application. Any new findings that suggest the evidence used to inform a policy or intervention has the potential to increase an inequality of a health outcome would warrant a rethinking of the policy or intervention. In this paper, we report such a finding, as we describe the relationship between SES and excess body weight in Victoria, by sex, with careful consideration of the definitions used to define excess body weight. Submitted: July 2012 Revision requested: October 2012 Accepted: December 2012 Correspondence to: Ms Alison Markwick, Health Intelligence Unit, Prevention and Population Health Branch, Department of Health, 50 Lonsdale Street, Melbourne, Victoria 3000; alison.markwick@health.vic.gov.au 32 AUSTRALIAN AND NEW ZEALAND JOURNAL OF PUBLIC HEALTH 2013 vol. 37 no. 1

2 Adult obesity and socioeconomic status Methods Data source Data were collected as part of the Victorian Population Health Survey (VPHS); an annual state-wide survey conducted by the Victorian Government Department of Health to provide information on the health and well-being of the population to inform policy and planning. 8 In 2008, for the first time, the sample size was increased to 34,168 to allow for estimates of health and well-being down to the level of the 79 local government areas (LGA) of the State of Victoria. The larger sample size also allowed for analyses to be conducted by SES. Sampling frame and stratification The VPHS is conducted by computer-assisted telephone interview in a randomly selected representative sample of males and females aged 18 years and over, who resided in private dwellings in Victoria. People who were homeless, itinerant, or in hospitals or other institutions were excluded from the survey. The sampling frame was an electronic listing of Victorian telephone exchange prefixes and localities, and random digit dialling was used to generate a sample of telephone numbers that formed the household sample. One person per household, aged 18 years or older, was randomly selected for interview, based on having the most recent birthday. The survey sample was stratified by LGA, with a target sample of 426 interviews per LGA. The total sample achieved was 34,168 completed interviews, including 808 in languages other than English. Ethical standards The Department of Health Human Research Ethics Committee approved the survey methodology and questionnaire content. Response rate The response rate, defined as the proportion of households where contact was made and an interview completed, was 64.9%. The proportion of respondents who did not report their total annual household income was 15.. Weighting The survey data were weighted to reflect the Victorian population, using the estimated resident population data within each LGA by sex and age group (18-24 years, years, years, years, years and 65 years and over). The data were also adjusted to reflect the probability of selection of the respondent and the number of telephone lines within the household. 8 Measurement of overweight, obesity and total annual household income Body mass index (BMI) is a simple and cost-effective index of weight-for-height that provides the most useful and widely-used population-based prevalence measure of obesity, for the purposes of monitoring and evaluating changes over time and making comparisons between populations. The World Health Organization (WHO) recommends that adults with a BMI of 25.0-<29.9 kg/m 2 be classified as overweight, while those with a BMI of greater than 30 kg/m 2 be classified as obese. 1 While there are well-documented sex and ethnic-specific differences suggesting that different ranges may be more appropriate, in the absence of an international consensus, the WHO continues to advocate their original ranges and many countries, such as Australia, continue to use these. Therefore, survey respondents were asked to report their height and weight. We calculated their BMI and classified their weight status according to the WHO guidelines. 1 Survey respondents were asked to indicate which of several income brackets their total annual household income in Australian dollars fell into: $20,000 or less; $20,001-$40,000; $40,001- $60,000; $60,001-$80,000; $80,001-$100,000; or greater than $100,000. Total annual household income included income before tax from all sources such as wages, social security payments, child support, and investments over the previous 12 months. Approximately 1 of survey respondents refused to divulge their total annual household incomes. Statistical analysis The survey data were analysed using the Stata statistical software package (StatCorp LP, College Station Texas, Version 10). Prevalence estimates by total annual household income were calculated for males and females who were categorised as being (a) overweight or obese, (b) overweight or (c) obese. Prevalence estimates were age-standardised to the 2006 Victorian census population, using the direct method of age-standardisation by 5-year age groups, to eliminate the impact of any differences in the age structure between income levels. Relative standard errors (RSE = standard error / point estimate x 100) were calculated to assess the reliability of the prevalence estimates. When RSEs were below 2, prevalence estimates were deemed to be reliable. Ordinary least squares linear regression of the logarithms of age-standardised prevalence estimates on income category was performed. We included interaction terms to test for effect modification. The statistical significance of the slope of the regression line was used to assess the SES gradient. Results Prevalence of overweight and/or obesity in adults Table 1 shows the prevalence of overweight (BMI of kg/m 2 ), obesity (BMI of 30 kg/m 2 or greater) and overweight and obesity combined (BMI of 25.0 kg/m 2 or greater), by total annual household income in Victoria in Figure 1 depicts the relationship between the prevalence of overweight, obesity, or overweight and obesity combined, by total annual household income in Victorian adults. There was a statistically significant positive association (reverse SES gradient) of the prevalence of overweight and SES. By contrast, there was a statistically significant negative association (typical SES gradient) of the prevalence of obesity and SES. When we combined the two 2013 vol. 37 no. 1 AUSTRALIAN AND NEW ZEALAND JOURNAL OF PUBLIC HEALTH 33

3 Markwick, Vaughan and Ansari categories of overweight and obesity, we no longer observed an SES gradient. Prevalence of overweight, by sex We observed a sex difference in the relationship of SES and overweight and confirmed the sex difference by including an interaction term, between sex and income category, in our initial multiple regression model. The interaction term was significant (p=0.006) thus justifying our analysis of the data separately for each sex. There was a statistically significant reverse gradient for males, where the prevalence of overweight increased with increasing SES (Figure 2). By contrast, there was no statistically significant SES gradient of overweight in females in either direction. Prevalence of obesity, by sex We did not observe a sex difference in the relationship of SES and obesity. There were statistically significant typical SES gradients of obesity, in both males and females, where the prevalence of obesity decreased with increasing SES (Figure 2). Prevalence of overweight and obesity, by sex When the overweight and obese weight categories were combined, there remained a statistically significant reverse SES gradient in males, but no SES gradient in females (Table 1). Discussion We investigated the relationship of overweight and obesity with SES, using the WHO recommended ranges for BMI as indicators of overweight and obesity, and total annual household income as the indicator of SES. We show that combining the categories of overweight and obesity into a single category masks important SES differences, while combining the data for males and females masks important sex differences. While excess body weight is treated as a continuum in terms of its risk to health, our findings show that the same cannot be assumed with regards to its relationship with SES. Moreover, we add to the body of literature by showing that in the Australian State of Victoria excess body weight in developed countries is also negatively associated with SES in females, but not necessarily males. 6,7 Table 1: Prevalence of overweight and obesity, by sex and total annual household income. Total annual household income Overweight (BMI = ) Obese (BMI 30.0) Overweight or obese (BMI 25.0) Males % 9 CI RSE % 9 CI RSE % 9 CI RSE Less than $20,000 $20,001 $40,000 $40,001 $60,000 $60,001 $80,000 $80,001 $100,000 More than $100,000 Don t know or refused to say Total adult males % 9% Females Less than $20,000 $20,001 $40,000 $40,001 $60,000 $60,001 $80,000 $80,001 $100,000 More than $100,000 Don t know or refused to say Total adult females % Persons Less than $20,000 $20,001 $40,000 $40,001 $60,000 $60,001 $80,000 $80,001 $100,000 More than $100,000 Don t know or refused to say Total adults Notes: 9 CI = 95 per cent confidence interval. RSE = relative standard error. Data are age-standardised to the 2006 Victorian population by 5-year age groups. 34 AUSTRALIAN AND NEW ZEALAND JOURNAL OF PUBLIC HEALTH 2013 vol. 37 no. 1

4 Adult obesity and socioeconomic status Figure 1: Prevalence of overweight and obesity, by total annual household income in Victorian adults Per cent (9 CI) Overweight Obese Overweight and obese Less than $20,000 $20,001 - $40,000 $40,001 - $60,000 $60,001 - $80,000 $80,001 - $100,000 More than $100,000 Total annual household income Figure 2: Prevalence of overweight and obesity, by sex and total annual household income Per cent (9 CI) Overweight males Overweight females Obese males Obese females Less than $20,000 $20,001 - $40,000 $40,001 - $60,000 $60,001 - $80,000 $80,001 - $100,000 More than $100,000 Total annual household income 2013 vol. 37 no. 1 AUSTRALIAN AND NEW ZEALAND JOURNAL OF PUBLIC HEALTH 35

5 Markwick, Vaughan and Ansari Our findings are consistent with the observation that the increase in the prevalence of unhealthy excess weight may be due to an increase in the proportion of persons with extremely high BMI rather than a shift of the entire BMI distribution. 9,10 Given that people of low SES disproportionately bear the burden of ill-health and constitute a relatively high proportion of the population, it follows that one would observe that obesity, rather than overweight, is associated with low SES. Zhang and Wang reported similar findings in the US, using measured BMI and the concentration index (CI) a summary measure of socioeconomic inequality where the cumulative proportion of the population, ranked by income from poorest to richest, was plotted against the cumulative proportion of the obese (defined as a BMI 30 kg/m 2 ) or overweight (defined as a BMI 25 kg/m 2 ). 11 They observed typical SES gradients in both obese males and females, a typical SES gradient for overweight females and a reverse SES gradient in overweight males. We also observed typical SES gradients in both obese males and females, and a reverse SES gradient in overweight males and overweight combined with obese males. However, the typical SES gradient in obese females disappeared when we included overweight females. It should be noted that Zhang and Wang s definition of overweight was equivalent to our definition of overweight combined with obesity. Zhang and Wang suggest that the use of different definitions of obesity may contribute to the variation in findings of SES gradients, and we concur. McLaren (2007) found that of 56 published studies of selfreported BMI by an income indicator of SES in males in developed countries, 16 found reverse SES gradients, 9 found typical SES gradients and 31 failed to find an SES gradient. 7 We reviewed these studies and found that of the 16 studies that reported reverse SES gradients, only 7 had comparable data and were able to be accessed electronically. Of these, the majority (6 studies or 8) used BMI as a continuous variable or used a BMI cut-off that included overweight males. By contrast, of the 9 studies that reported typical SES gradients, for which 6 of the studies contained comparable data that could be accessed electronically, the majority (5 studies or 8) used a BMI cut-off of 30 kg/m 2 or greater. We believe that the definition of obesity, employed in studies evaluating SES gradients, may account for some of the variation in findings, particularly in males. If the term obesity is used loosely to include males with a BMI of between 25.0 and 29.9 kg/m 2, that is, those who are overweight rather than obese, it is more likely that a reverse or no SES gradient between body size and income may be detected. By contrast, if an appropriate BMI cut-off of 30 kg/ m 2 is used, a typical SES gradient is more likely to be found. This implies that the type of SES gradient observed may depend on the relative proportion of overweight to obese males in any given study. Where there is a predominance of overweight males, one may be more likely to observe a reverse SES gradient. Conversely, where there is a predominance of obese males, a typical SES gradient is more likely, and at some point in between it is likely that no SES gradient will be observed. Choice of SES indicator It is quite possible that the disparate findings in the relationship between SES and obesity in males may also reflect the choice of SES indicator as well as the choice of obesity indicator. For example, the 2003 Health Survey of England found typical SES gradients of obesity in females regardless of the SES indicator used. 12 By contrast, typical SES gradients for males were only found when occupation-based and qualification-based but not income-based or area-based measures of SES were used. 12 We used an incomebased SES indicator and found a typical SES gradient of obesity in males, which is not consistent with the UK findings. However, while the UK findings did not find an SES gradient for obesity (BMI 30 kg/m 2 ) in males by equivalised household income, they did find a reverse gradient when they combined overweight with obesity (BMI 25 kg/m 2 ), consistent with our findings. In Australia, area-based indicators of SES are commonly used in the absence of individual level data. Glover et al. 13 using population quintiles of the 1996 census Index of Relative Socio-Economic Disadvantage (IRSD) and self-reported BMI categorised using the WHO cut-offs, reported a statistical association of unhealthy weight with one or more IRSD quintiles, but no gradient, in either males or females. They categorised unhealthy weight as being overweight or obese. 13 The lack of observation of a gradient could reflect the use of a different indicator of SES as well as the unit of analysis being a geographic area rather than the individual. By contrast, Cameron et al. 14 reported significant typical SES gradients in both males and females using highest level of educational attainment as the SES indicator and the WHO BMI cut-off for obesity ( 30 kg/m 2 ). While this was consistent with our findings, they also used an income-based SES indicator (total annual household income) and found a reverse SES gradient for females but no SES gradient for males. This is at odds with our findings. However, their study had a very low response rate, with only 49. of eligible households participating in the household interview and only just over half of these (55.) participating in the biomedical examination where they had their BMI measured. 15 Also, the proportion who agreed to provide income information was not reported. Moreover, the study was conducted between 1999 and 2000 and it is possible that the situation has changed as obesity rates continue to rise. Our findings provide a more recent and representative snapshot of the relationship between unhealthy weight and SES in Victoria. Why are females more susceptible to the typical SES gradient of obesity than males? Females of low SES may be more vulnerable to developing obesity because they are disproportionately subject to social and environmental pressures, such as discrimination in employment and income, less physical activity, pregnancy, responsibility for household budget and organising meals, pressure from mass media, marketing and societal norms, and resulting low self-esteem. 16 Females and people of higher SES are also significantly more likely to perceive and identify overweight than their male or lower 36 AUSTRALIAN AND NEW ZEALAND JOURNAL OF PUBLIC HEALTH 2013 vol. 37 no. 1

6 Adult obesity and socioeconomic status SES counterparts. 17 Females have been shown to be more sensitive to perceived overweight and may face greater social pressures to conform to ideals of body image than males. 18 Social mobility may also contribute to the SES gradient of obesity in females, as upwardly socially mobile females have been shown to be much thinner than downwardly socially mobile females, or females that did not change their SES. No such relationship was observed for males. 19 Clear and consistent stigmatisation of overweight females, but not males, is well documented in three areas of life: employment, education and health care. 20 Why is there a reverse SES gradient for overweight males? A possible explanation is that low SES males are more likely to work in manual professions, which require a moderate level of physical activity, and therefore engage in higher levels of physical activity through incidental work-related physical activity. We asked respondents whether they engaged in physically demanding work in their paid employment. We found statistically significant typical SES gradients in both overweight, but not obese, males and females where the proportion of persons engaging in physically demanding work declined with increasing total annual household income (data not reported). However, this type of physical activity was not captured as part of the data used to estimate overall weekly physical activity levels. Therefore, we cannot confirm whether these findings translated into higher overall physical activity levels. Another explanation lies in the evidence that there is less social pressure on males to maintain a healthy weight and so being a little overweight is more likely to be tolerated. Males of higher SES are more likely to have the knowledge, inclination and capacity to address the matter at that point. The reverse SES gradient seen in overweight males may reflect the fact that higher occupation status tends to be associated with lower levels of physical activity, up to the point when the weight is perceived to be excessive and corrective action is taken. Main strengths and limitations of study A main strength of this study is the large randomly selected sample size that is representative of all local government areas in Victoria. Another strength is the collection of total annual household income, using a wide range from less than $20,000 to more than $100,000. This study has a number of limitations. It is based on crosssectional data and cannot make any inferences about causality. Whether being of low SES predisposes one to obesity, or being obese predisposes someone to being of low SES, cannot be determined. The literature provides evidence for both mechanisms. We used self-reported height and weight to calculate BMI. It is well established that individuals tend to over-report height and under-report weight, resulting in the under-estimation of the absolute prevalence of overweight and obesity. 21 However, self-reported BMI is still useful for epidemiological and monitoring purposes, as it offers an affordable and practical method of evaluating population body size for the purposes of making relative comparisons between populations and over time. It is possible, however, that self-reporting of height and weight could vary across different SES groups, and that this may contribute to the SES gradient. Higher SES groups have been shown to perceive their body size more accurately and this could result in more accurate reporting. 22 However, our results concur with the findings of Zhang and Wang, and in their study BMI was measured, not self-reported. Household size is an important potential confounding variable, as the smaller the household, the greater the resources available to each individual. The results we report in this paper were not adjusted for household size, as information on total annual household income was captured in ranges, not absolute amounts. However, we used the mid-point of each income bracket as a proxy of absolute household income and divided by the total number of people normally resident in the household, applying the Organisation for Economic Cooperation and Development modified scale that assigns a weighting of 0.5 to second and subsequent household members and a weighting of 0.3 to children. 23 We reproduced our results (data not shown) indicating that household size was not an important determinant and did not influence our findings. Appropriateness of the WHO BMI ranges for females of reproductive age is currently being debated, with evidence showing that the BMI cut-off value of 30 kg/m 2 failed to identify almost half a sample of females who met the criteria for obesity by per cent body fat. 24 The picture is further complicated by ethnic differences. To date, no consensus has been reached on sex and ethnic-specific BMI ranges. Therefore, the WHO BMI ranges, as they currently stand, continue to provide the most commonly used public health indicator of overweight and obesity used across the world to report and monitor the health of populations and act as a trigger for action. Conclusions Our data show that obesity, when defined as a BMI of 30 kg/ m 2, is associated with socioeconomic disadvantage in both males and females. By contrast, overweight, when defined as a BMI of 25 to 29.9 kg/m 2, is associated with affluence in males and does not appear to be associated with SES in females. If policies and interventions seek to reduce the burden of chronic disease due to unhealthy weight, without increasing social inequalities in health, we should refrain from the practice of combining overweight and obesity into a single category, because this will impair our ability to appropriately identify and target high-risk communities. Acknowledgements We thank Dr Leonard S. Piers of the Health Intelligence Unit, Department of Health, for reviewing and critiquing this work. The views expressed in this article are those of the authors and do not necessarily represent those of the Victorian Department of Health or the Victorian Government vol. 37 no. 1 AUSTRALIAN AND NEW ZEALAND JOURNAL OF PUBLIC HEALTH 37

7 Markwick, Vaughan and Ansari References: 1. World Health Organisation. Obesity, Preventing and Managing the Global Epidemic. Report of a WHO Consultation on Obesity. Geneva (CHE): WHO; Olshansky SJ, Passaro DJ, Hershow RC, Layden J, Carnes BA, Brody J, et al. A potential decline in life expectancy in the United States in the 21st century. N Engl J Med. 2005;352(11): Australian Institute of Health and Welfare. National Healthcare Agreement: PB 03-By 2017, Increase by Five Percentage Points the Proportion of Australian Adults and Children at a Healthy Body Weight, Over the 2009 Baseline (Baseline Specification), 2012 [Internet]. Canberra (AUST): Commonwealth Of Australia; 2012 [Cited 2012 Nov 15]. Available from: index.phtml/itemid/ Australian Bureau of Statistics Overweight and Obesity in Adults in Australia: A Snapshot [Internet].Canberra (AUST): ABS; 2011 [Cited 2012 Nov 13]. Available from: Ausstats/subscriber.nsf/0/7DC7186F4A9950DECA25789C0023DCEF/$Fi le/ _ pdf 5. Tasmanian Department of Health and Human Services. National Partnership Agreement on Preventive Health - Performance Benchmarks [Internet]. Hobart (AUST): State Government of Tasmania; 2011 [Cited 2012 Nov 15]. Available from: data/assets/pdf_ file/0008/83690/ _npaph_3_performance_benchmarks.pdf 6. Sobal J, Stunkard A. Socioeconomic status and obesity: A review of the literature. Psychol Bull. 1989;105(2): McLaren L. Socioeconomic status and obesity. Epidemiol Rev. 2007;29:29 8. Victorian Department of Health. Victorian Population Health Survey [Internet]. Melbourne (AUST): State Government of Victoria; 2008 [Cited 2012 Jun 01]. Available from: 9. Beyerlein A, Toschke AM, von Kries R. Risk factors for childhood overweight: shift of the mean body mass index and shift of the upper percentiles: results from a cross-sectional study. Int J Obes. 2010;34: Terry MB, Wei Y, Esserman D. Maternal, birth, and early life influences on adult body size in women. Am J Epidemiol. 2007;166: Zhang Q, Wang Y. Socioeconomic inequality of obesity in the United States: do gender, age, and ethnicity matter? Soc Sci Med. 2004;58: Sproston K, Primatesta P, editors. Risk Factors for Cardiovascular Disease. In: Health Survey for England Volume 2. London (UK): The Stationery Office; Glover JD, Hetzel DM, Tennant SK. The socioeconomic gradient and chronic illness and associated risk factors in Australia. Aust N Z Health Policy. 2004;1: Cameron A, Welborn T, Zimmet P, Dunstan D, Owen N, Salmon J, et al. Overweight and obesity in Australia: the Australian Diabetes, Obesity and Lifestyle Study (AusDiab). Med J Aust. 2003;178: Dunstan D, Zimmet P, Welborn T, Cameron A, Shaw J, de Courten M, et al. The Australian diabetes, obesity and lifestyle study (AusDiab) - methods and response rates. Diabetes Res Clin Pract. 2002;57: Robertson A, Lobstein T, Knai C. Obesity and Socio-economic Groups in Europe: Evidence Review and Implications for Action [Internet]. Brussels: European Commission; 2007 [Cited 2012 Feb 04]. Available from: ec.europa.eu/health/ph_determinants/life_style/nutrition/documents/ ev _rep_en.pdf 17. Paeratakul S, White M, Williamson D, Ryan D, Bray G. Sex, race / ethnicity, socioeconomic status, and BMI in relation to self-perception of overweight. Obes Res. 2002;10(5.): Sanchez-Vaznaugh E, Kawachi I, Subramanian S, Sanchez B, Acevedo-Garcia D. Do socioeconomic gradients in body mass index vary by race/ethnicity, gender, and birthplace? Am J Epidemiol. 2009;169: Wardle J, Robb K, Johnson F, Griffith J. Socioeconomic variation in attitudes to eating and weight in female adolescents. Health Psychol. 2004;23(3): Puhl R, Brownell K. Bias, discrimination, and obesity. Obes Res. 2001;9(12): McAdams M, Van Dam R, Hu F. Comparison of self-reported and measured BMI as correlates of disease markers in U.S. adults. Obesity. 2007;15(1): Wardle J, Griffiths J. Socioeconomic status and weight control practices in British adults. J Epidemiol Community Health. 2001;55(3): Organisation for Economic Co-operation and Development. What are Equivalence Scales? In: Income Distribution and Poverty: Data, Figures, Methods and Concepts [Internet]. Paris (FRA): OECD; 2011 [Cited 2012 Apr 10]. Available from: Rahman M, Berenson A. Accuracy of current body mass index obesity classification for white, black, and Hispanic reproductive-age women. Obstet Gynecol. 2010;115(5): AUSTRALIAN AND NEW ZEALAND JOURNAL OF PUBLIC HEALTH 2013 vol. 37 no. 1