The obesity epidemic is progressing in the United Kingdom (UK) [1, 2]. Forecasting obesity prevalence among the general population in the UK from current trends, the Foresight Obesity project suggested that 60% of men and 50% of women will be obese by 2050. The findings also suggest that social class differences in obesity by socioeconomic position (SEP) may widen with time [1].

Obesity is a central contributor to cardiovascular disease, being associated with hypertension, hypercholesterolemia, and coronary heart disease [3]. Obesity is also a predictor of several other diseases of population health importance [4, 5], including diabetes mellitus [6], cancer [7–9], stroke [10], and depression [11], among others [4]. Decreased life expectancy and excess mortality have also been demonstrated at both extremes of body mass index (BMI) [12, 13].

Low SEP is a well-documented determinant of poor health among diverse populations. There is a large literature assessing socioeconomic inequalities in several health indicators in the UK, including socioeconomic differences in heart disease, chronic bronchitis, smoking, diet, exercise, self-rated health, and overall mortality [14, 15] between rich and poor. A recently published comprehensive review about health inequalities in England highlighted SEP inequalities in morbidity, self-reported health, psychopathology, accidental injury, and mortality [16]. Several studies have also suggested that health inequalities by SEP may be widening in the UK, including life expectancy and mortality rates between the early 1980s and 2000s [16, 17].

Of particular importance here is the well-documented inverse relation between SEP and obesity risk--several studies have suggested socioeconomic disparities in adult obesity in the UK, with the poor at higher risk [18–22]. A recent data briefing from the UK's National Obesity Observatory demonstrated consistent inequalities in obesity over the last decade by occupational social class (OSC) among adults, with evidence of increasing disparities among both men and women [2].

Although there have been several published reviews about the relation between socioeconomic status and obesity risk that have included UK data [23, 24], to our knowledge, there has been no attempt to systematically appraise or synthesize the literature specific to this context. Our review was limited to the UK for several reasons: First, we were interested in understanding mechanisms that underpin SEP inequalities in the UK. As national health systems may influence access to health services, and may also determine the focus placed on prevention within countries, generalizing across countries may not be sensible. Second, there is a correlation between ethnicity and SEP in high-income countries, and members of ethnic minority groups have been shown to have differential risk for obesity than whites [25–29]. Countries with differing ethnic minority populations may therefore feature different relations between SEP and obesity, precluding generalization across countries. A nationally-focused review about socioeconomic differences in obesity risk in the UK is therefore warranted, as international reviews may lack the focus necessary to draw UK-specific inference that is useful for policy purposes or to direct further research to better understand the contribution of local context.

In this systematic review, we assessed the extant peer-reviewed literature published in the past 30 years about socioeconomic disparities in adult obesity in the UK. Summarizing important differences in the prevalence and determinants of obesity by different indicators of SEP in the UK, we attempted to isolate key indicators of socioeconomic position that may influence obesity risk. Moreover, we address generalizable themes and explore methodological limitations to the available literature.

We reviewed the peer-reviewed literature published between 1st January, 1980 and 8th March, 2010. Our review was limited to this period so as to reflect current thinking regarding the relation between SEP and health. We identified the literature reviewed through the MEDLINE database using the "http://pubmed.gov" interface. MeSH search terms "Obesity" and "Great Britain" were used to search for English-language articles published in the peer-reviewed literature. The MeSH term "Great Britain" includes any papers indexed with the following tags: "England", "Scotland", "Wales", "Northern Ireland", and "United Kingdom". All queries were carried out by the primary author during the month of March, 2010. A flow chart reporting studies excluded at each stage in the review process is shown in Figure 1.

To be included in the review, studies had to show evidence of having done each of the following:

• Considered differences in outcomes (e.g., obesity prevalence, mean BMI, etc.) by at least one defined measure of SEP, and described attribution of SEP measures among respondents

• Described the method used to define obesity, including metric of interest, and threshold for obesity utilized in analysis

• Conducted a direct empiric analysis of differences in obesity outcome by measure of SEP

The primary author extracted the following information from each paper: definition of obesity; socioeconomic position measure(s); population and setting; sample and methods; covariates included in final models; and findings and conclusions. We organized our findings by study design, considering first longitudinal analyses of the relation between SEP indicators in earlier life and obesity risk at a later point during the life course, and then moving to cross-sectional analyses of the relation between SEP indicators and obesity.

Within each study design, we first considered the relationship between individual and family-level measures of SEP and obesity. As markers of SEP at the individual or family levels, studies employed the following measures among individuals or heads of households: occupational social class (a measure of social class by employment type), educational attainment, salary scale, income, receipt of government aid, access to various resources, and/or employment history, among others. We also considered studies about the relationship between area-level measures of SEP and obesity. To assess area-level SEP, studies employed the following measures: various deprivation indices, proportions of the population by context in manual occupations, and/or proportions of the population by context renting housing from the local authority, among others. The outcomes considered in these studies included: BMI, BMI cutoffs for overweight or obesity (cutoffs employed varied by study), waist/hip ratio, waist circumference, and/or weight/height ratio, among others.

Summary measures considered included differences in mean BMI, mean waist/hip ratio, or mean waist circumference, and/or risk ratios or odds ratios of overweight or obesity (employing BMI, waist/hip ratio, or waist circumference cutoffs), among others. Given the heterogeneity in area-level and individual/household-level measures of SEP employed, as well as the multiplicity of metrics of obesity in the literature reviewed here, a meta-analysis of the results was not pursued.

Additional file 1: Table S1 features a detailed review (including SEP indicator used, definition of obesity, setting and population, sample and methods, covariates included in final models, and findings and conclusions) of each study. Our original search yielded 1189 articles, 233 of which were judged to consider the relation between SEP and obesity in the UK after screening by title. Upon screening by abstract for empirical articles set in the UK, we were left with 102 articles. After reading the remaining articles, another 54 were discarded because they did not meet the specified inclusion criteria. Reference lists from the remaining 48 articles were searched, and yielded a further 10 articles which fulfilled the inclusion criteria, leaving a total of 58 articles. Finally, 23 articles did not include outcome measure among respondents older than 18 years, and were excluded, yielding a total of 35 articles reviewed here.

Studies in this review featured two empirical designs: 20 studies were longitudinal analyses, and the remaining 15 were cross-sectional. Only four studies included socioeconomic measures collected at multiple levels (area-level, household/individual-level), and only two of these studies utilized multilevel modeling approaches. None of these studies utilized systems modeling approaches in analysis.

There were 19 studies that analyzed representative data from at least one country in the United Kingdom: 12 studies reported on data from Wales; 14 reported on data from Scotland; and 19 studies reported on data from England. There were no studies that considered data from Northern Ireland. The remaining 16 of the studies we reviewed analyzed data from regional datasets from localities throughout the UK (London, Newcastle, etc.).

A systematic review of the peer-reviewed literature about socioeconomic inequalities in adult obesity in the UK published between 1980 and 2010 found that socioeconomic indicators of low SEP throughout the life course as well as in cross-sectional analyses, including head-of-household OSC at birth and during childhood, earlier adulthood OSC, current OSC, educational attainment, and area-level deprivation were reliably associated with higher obesity risk in the UK. Notably, several indicators, including low head-of-household childhood OSC and low adulthood OSC, were found to be more strongly associated with obesity among women than among men. There may also be ethnic differences in the relation between SEP and obesity risk.

This is the first systematic review, of which we are aware, to consider the relation between SEP and obesity in the UK. However, our findings are supported by other systematic reviews about socioeconomic inequalities in obesity in high-income contexts that have shown an inverse relation between SEP and obesity risk [23, 24, 60]. Our findings are also supported by the conceptual literature about socioeconomic inequalities [61]. In their work on fundamental causes, Link and Phelan posit that higher SEP will always predict better health because SEP, through access to more knowledge, money, power, social connectedness, and prestige, affords access to resources that can optimize health across societies in all times [61]. In a society, such as the UK, where cardiovascular disease is responsible for a third of all deaths [62], it is plausible, then, that lower SEP should predict higher risk for obesity, a critical modifiable risk factor for cardiovascular disease.

There are several mechanisms that may mediate the relation between SEP and obesity risk in high-income countries. First, education is a principle component of SEP, predicting both income and social class. Independently of material pathways, however, education, itself, may also predict obesity risk via access to health information and perceived agency [33]. Education portends health literacy, as less-educated individuals may lack the numeracy required to understand health advice from health providers [63] or the literacy required to access health information available in other media. The resultant lack of information among the less-educated may then shape food and physical activity choices, as has been demonstrated in findings from the Low Income Diet and Nutrition Survey, which showed that those without educational qualifications had lower fruit and vegetable consumption and higher consumption of energy-dense foods as compared with those with even the educational lowest qualifications [64]. Moreover, less educated individuals may lack the confidence or perceived agency to improve their health. For example, a recent study of 1,967 women aged 18-34 in Scania, Sweden demonstrated the constellation of low education and behaviors portraying low locus of control among overweight and obese women relative to their counterparts who were underweight or had normal weight [65].

Higher income and social class also operate to mitigate obesity risk. Most directly, income, as well as social class (highly-correlated with income) may promote a healthier diet via direct access to healthier food options [64]. These factors may also protect against obesity via more consistent access to food. Food insecurity, defined as limited or uncertain availability of nutritionally adequate and safe foods [66], may promote obesity by incentivizing binge-eating--as food insecure individuals may be uncertain about the availability of their next meals, they may binge on meals when they are, in fact, available. Aggregated over time, this compensatory behavior can increase obesity risk [67, 68].

Income and social class may also shape residential decisions. Area-level SEP may predict obesity risk in important ways [44], and neighborhoods may shape obesity risk via several mechanisms [69]: Low-income neighborhoods may have less green space and lower walkability, which may discourage physical activity [69]. Furthermore, low-income neighborhoods may limit access to healthy foods, limiting the quality of diets among residents [69]. Low-income neighborhoods are also characterized by lower social capital, a measure of inter-member trust and support that is influenced by the degree of crime, safety, and disorder in a context. In that vein, a recent study by Poortinga and colleagues demonstrated that low social capital may increase obesity risk [70], suggesting that even beyond access to material resources, characteristics of communities in low-income neighborhoods may influence obesity.

The distinct social history of the cohort of UK adults considered in a large number of the studies reviewed here may also be important. The World War II and reconstruction eras, into which many of these adults were born, were turbulent economic times in the UK [71]. Between 1939 and 1955, essential food supplies, clothing, and household products were rationed by the British government to bolster the war effort and accelerate post-war reconstruction [71]. A consequence of this policy, however, was the accentuation of class differences in food access, as the wealthy were able to supplement their rations via other means [71].

Several studies have suggested that the macronutrient environment in early development may be particular important in determining obesity risk in later life [72–74]--and that food scarcity during development may predict obesity in later life [74]. Indeed, findings from many of the studies we reviewed here suggest that exposure to low socioeconomic position in childhood may increase risk for adult obesity [18–22, 32–41]. As a substantial proportion of the adults sampled in the studies we reviewed here were born during the era of government food rationing, it is plausible that some of the adulthood differences in obesity risk by SEP observed here may reflect, in part, intrauterine or early childhood macronutrient scarcity, particularly among children from low SEP households who maintained low SEP in adulthood.

There are several limitations that should be considered when interpreting the findings reported here. First, because our inclusion criteria limited the studies reviewed here to those published in the peer-reviewed literature, the inferences we have drawn may be subject to a publication bias. Although we used relatively permissive inclusion criteria, and included studies analyzing many of the largest health surveys in the UK, our findings may not accurately reflect current knowledge about SEP and obesity in the UK. Second, our search strategy included a query of only one database, and therefore, it is plausible that some of the literature about the relation between SEP and obesity risk may not have been represented in our findings. However, a detailed query of the citations of all studies found via our initial search was conducted to minimize this possibility. Moreover, we were interested in reviewing the public health and medical literatures. In this light, a recent study of the utility of the four most prominent biomedical databases demonstrated that MEDLINE was the optimal tool for searches of the biomedical literature [93]. Third, there was substantial overlap with respect to the health surveys analyzed in the studies we reviewed, which may limit the breadth of our findings. However, this is a limitation imposed by the literature itself and was unavoidable. Fourth, our findings were organized by a data type, and by SEP indicator. This organizational scheme may have, in part, shaped the inferences drawn here. Fifth, our findings were limited to studies about socioeconomic disparities in obesity risk among adults in one European country. It would therefore be inappropriate to generalize our findings to other contexts.

Our systematic review of the peer-reviewed literature between 1980 and 2010 demonstrated considerable inequalities in obesity by SEP in the UK. However, there remain several limitations to our understanding of the relation between SEP and obesity in the UK. Considering these limitations, we suggest that investigators interested in SEP disparities in obesity in the UK pursue three avenues of inquiry. First, future studies about the relation between individual-level SEP and obesity might operationalize individual-level SEP using common measures of SEP other than OSC, including educational attainment and/or income. Second, the conceptualization and analysis of future studies in this area should consider multilevel and complex systems approaches that account for SEP influences at multiple levels, including the individual, household, and area levels, on risk for obesity in this context. Third, future work may explicitly examine differences in the relation between SEP and obesity by ethnicity in the UK, as current work has suggested that SEP may interact with ethnicity to influence obesity in important ways.