Unevenly distributed: a systematic review of the health literature about socioeconomic inequalities in adult obesity in the United Kingdom

Of the 35 articles reviewed here, 17 showed differences in the relation between SEP and obesity risk by gender [18, 21, 31, 32, 34, 38‐40, 42‐45, 47, 54, 55, 57, 58]. Overwhelmingly, the literature suggests that SEP measures are more strongly and reliably associated with obesity among women than among men (as demonstrated by 13 of 17 studies) [18, 21, 31, 32, 39, 40, 42, 44, 47, 54, 55, 57, 58].

As discussed above, studies found consistent differences in both the relations between childhood OSC and risk for obesity in adulthood, as well as adulthood OSC and concurrent risk for obesity by gender. Moreover, one study found that although area-level deprivation (Townsend Material Deprivation Score [43]) was not associated with BMI increase among men or overall, it was associated with BMI increase among women [44].

The finding that SEP may be more strongly inversely associated with obesity risk among women than among men is consistent with other systematic reviews of the literature about socioeconomic inequalities in obesity in high-income contexts [24, 60]. While it remains unclear why SEP may be more strongly and reliably associated with obesity risk among women, this dimorphism has a plausible explanation. Aside from one study [44], OSC was the socioeconomic measure employed in all of the other studies that found gender differences in the relation between SEP and obesity [18, 21, 31, 32, 34, 38‐40, 42, 43, 45, 47, 54, 55, 57, 58]. The literature about socioeconomic measures in the UK suggests that women in the same occupations, and therefore the same OSCs, may receive lower remuneration than men [75, 76]. Women workers may also be concentrated into fewer and lower-paid occupations per OSC classification than men [77]. In this way, OSCs may not be comparable across genders, and lower OSC categories among women may reflect substantially more disadvantage relative to their male counterparts. This difference in SEP indicated by OSC by gender, therefore, may in part explain the stronger relationship between OSC and obesity risk among women as compared to men in this literature.

While the present review draws attention to important socioeconomic gradients in obesity risk in the UK, there are several limitations to the present literature that challenge our understanding of the relation between SEP and obesity risk among adults in this context: 1) the overreliance on occupational social class (OSC) as the principle socioeconomic measure in extant studies, 2) few studies (three out of 35) that have simultaneously considered SEP indicators at both the area-level and the individual/household-level, 3) few studies (two out of 35) have utilized multilevel or systems modeling techniques to assess the potential for socioeconomic influences on obesity at multiple levels, and 4) a paucity of studies (one out of 35) that have utilized ethnically diverse datasets, and/or assessed differences in the relation between SEP indicators and obesity by ethnic group.

The first methodological limitation to the extant literature is the overreliance on occupational social class (OSC) as a measure of SEP in studies concerned with inequalities in obesity in the UK. To frame this limitation, of the studies reviewed here, only 11 out of 35 considered SEP indicators other than OSC. And among those 11 studies, there were nearly twenty other measures of SEP considered. The next most utilized SEP indicator was education, which was only considered in four (as compared to 25) studies reviewed here. Taken together, these findings suggest that our understanding of SEP inequalities in obesity in the UK is heavily dependent on the OSC indicator, and that there are relatively few comparably well-studied indicators upon which to base our understanding of SEP disparities in obesity in the UK.

The Occupational Social Class indicator was developed in 1913 by British Registrar General THC Stevenson, and has regularly been collected in UK datasets since that time [20, 60, 78]. As termed by Stevenson, the indicator was meant to capture "standing within the community" or "culture" [78, 79]. Shown to be reliably predictive of morbidity and mortality [20, 79, 80], similar indicators have been adapted in several other European countries [77].

There are several deficiencies to the OSC as a measure of SEP (for review, see Krieger and colleagues [77]), because of which, the literature about SEP disparities in adult obesity in the UK is challenged by an overreliance on the indicator. First, there may be considerable heterogeneity in exposure to poverty and potentially pathogenic occupational exposures by ethnicity and gender within a given OSC [77]. For example, as noted above, women, along with ethnic minorities in the same occupations have been shown to receive lower remuneration than men and whites in the UK, even after accounting for education and work experience [77, 80]. Moreover, evidence in this context has suggested that women workers may be concentrated into fewer and lower-paid occupations per OSC classification than men [79]. Second, the OSC may not accurately identify the SEP of individuals outside of the market labor force, such as the unemployed, retired adults, children, and individuals employed in informal sectors, such as homemakers [81]. Although head-of-household OSC measures may be used as proxies for measuring SEP among individuals who fall into the above classifications, these proxies do not account for differences in family structure and/or dependency in relation to the head-of-household. Third, this measure may not be comparable across economic spatial or temporal contexts, as distributions of wealth, prestige, and exposure to potentially pathogenic occupational hazards may be different across occupations in different spatial and temporal contexts. This heterogeneity may therefore limit comparisons of the relations between OSC and health metrics across contexts in space and time.

The second two limitations to our understanding of SEP inequalities in adult obesity, that only three studies that have simultaneously considered SEP indicators at both the area-level and the individual/household-level, and that only two have utilized multilevel techniques (none that have used systems modeling techniques) to assess the potential for socioeconomic influences on obesity at multiple levels in the UK, are of fundamental importance. The notion that individuals may interact, and thus be influenced by, their ecological contexts is foundational in population science research [82‐85]. Studies concerned with SEP inequalities in adult obesity which only consider variation in obesity using measures of SEP at the individual or household level (29 of 35 studies reviewed) may not appropriately account for the etiologic impact of ecological poverty on obesity, and therefore may yield an incomplete assessment of the association between SEP and obesity. Rather, studies that simultaneously consider both individual and area-level factors as determinants of outcomes are most appropriate, given the following three considerations: First, individuals interact with their ecological contexts, and are therefore potentially influenced by them [82‐85]. Second, area-level SEP variables may be poor proxies for individual-level SEP. And third, quantifying the direct and indirect contributions of area-level SEP indicators to outcomes of interest in epidemiologic analyses that do not include individual-level indicators is challenging. Over the past several years, therefore, epidemiologists have begun to conceptualize and analyze etiologic models of disease from a multilevel perspective [86], which has presented a movement away from traditional models focusing exclusively on indicators at the individual-level, or proxies thereof [87]. Accounting for clusters within data nested at multiple levels of aggregation, multilevel models, allow the researcher to estimate mutually-adjusted exposure effects across levels of influence [82]. This approach to etiologic conceptualization and analysis has allowed investigators to consider how characteristics at several levels of influence--individuals, households, neighborhoods, cities, countries, and societies--may produce, individually and collectively, health and disease [86].

Emerging from this paradigm, as well as responding to a need for novel approaches to epidemiologic analysis, and the limitations of deterministic modeling, complex systems approaches utilize stochastic modeling techniques, allowing researchers to capture dynamic, bi-directional, and relational interactions between "exposures" and "outcomes" at several levels of influence [86]. Therefore, these approaches may be ideal for investigating the etiology and consequences of SEP inequalities in obesity in high-income contexts, such as the UK. In the absence of collective study of SEP measures at multiple levels of influence using multilevel or complex systems tools, our understanding of SEP disparities in obesity and their etiologies remains limited.

The fourth limitation to our understanding of the relation between SEP and obesity in the UK is a paucity of studies that have of utilized ethnically diverse datasets, and/or assessed differences in the relation between SEP indicators and obesity by ethnic group. There was only one study [47] concerned with differences in the relation between SEP and obesity by ethnic group, and this study found, as discussed above, potentially important differences in the relation between SEP and obesity by ethnic group. Many longitudinal studies about SEP disparities in obesity (8 of 20) utilized data from the 1946 and 1958 British birth cohorts, which do not adequately represent ethnic minorities in the UK of the 21st century [88].

Ethnic minorities are a large and growing subpopulation in the UK. Data from the most recent UK census (2001) [89] indicates that ethnic minority groups in the UK comprise over 8% of the total population, with about 4.6 million ethnic minority individuals in the UK. There are important socioeconomic differences between the ethnic minority and white UK populations. Ethnic minorities tend to be of lower SEP than their white counterparts. For example, Pakistani and Bangladeshi groups have the lowest proportions in "managerial and professional occupations" OSCs, and Bangladeshis and Black Africans in the UK have the highest proportions of children eligible for free school meals [90]. Ethnic minorities are more likely to be unemployed, and to have no educational qualifications [91]. Disparities in the healthcare experiences of ethnic minorities and whites have also been documented. For example, ethnic minorities are less likely to report positive experiences with healthcare providers compared to whites [90, 92].

Given the size of the ethnic minority population in the UK, as well as the substantial demographic differences between these populations and the general population in this context, it is plausible, as supported by the extant work [47], that there are important differences in the relation between SEP and obesity by ethnic group. The paucity of studies that have considered this relation, or have used ethnically-representative datasets presents a limitation to our understanding of inequalities in obesity, as it limits our understanding of how ethnicity and SEP may interact to determine obesity risk.