Independent and combined associations between fast-food outlet exposure and genetic risk for obesity: a population-based, cross-sectional study in the UK

As a modifiable factor associated with elevated risk of type 2 diabetes [1], cardiovascular disease [2], some cancers [3], and all-cause mortality [4], obesity is a global public health priority. In England, high body weight is one of the leading drivers of mortality and morbidity [5], and obesity costs the National Health Service £6.1bn per year [6]. One of the greatest challenges to addressing this public health problem at a population level is that obesity is a complex, multifactorial condition, including biological and social determinants.

Among biological determinants, genetic epidemiological studies have firmly established the role of genetics in the development of obesity. Genome-wide association studies (GWAS) have been able to identify individual genetic loci associated with obesity and common metabolic diseases [7]. For example, single-nucleotide polymorphisms (SNPs) in the FTO gene are associated with higher risk of obesity and type 2 diabetes [8], while SNPs in the MC4R gene are associated with increased fat mass and obesity risk [9]. The most recent GWAS have observed SNPs at 97 loci that can collectively indicate an individual’s genetic predisposition to obesity [10], although the level of variance explained is typically less than 2% [11].

The rise of obesity in the last 40 years has also been attributed to changing lifestyles [12], for example, declining physical activity levels and diets increasingly composed of processed, energy-dense foods, which can promote weight gain [13]. Food outlets selling these types of food, typically fast-food outlets, have become ubiquitous in neighbourhood environments [14, 15], and the potential of such access to influence dietary choices and long-term weight status is increasingly recognised [16]. Although some studies report contradictory or null findings, a growing number of cross-sectional and longitudinal studies in the UK and USA indicate that exposure to higher densities of fast-food outlets is associated with higher intake of energy-dense fast foods, higher body weight, and greater odds of obesity [17‐22]. However, none of these studies have been able to account for individual-level genetic predisposition to obesity.

Moreover, while there is strong evidence for both genetic and environmental drivers of obesity, to date there is little understanding of how these two types of determinants may interact. FTO, MC4R, and other SNPs have been associated with appetitive phenotypes including disinhibited eating and preferences for high-fat or energy-dense foods [23], which may be enabled in neighbourhood environments with higher densities of fast-food outlets. Therefore, we investigated the associations of genetic obesity predisposition and objectively-measured neighbourhood fast-food outlet exposure, independently and in combination, with body weight, in a population-based sample of UK adults. Motivated to understand a possible interaction, we also explored how associations between fast-food outlet exposure and body weight might be modified by genetic susceptibility to obesity.

In a sample of nearly 11,000 UK adults, we found that genetic predisposition to obesity and neighbourhood fast-food outlet exposure were independently associated with higher BMI and risk of obesity. We did not find evidence that the association of fast-food outlet exposure with BMI, overweight, or obesity differed between the low and high genetic risk groups. In combination, we observed a nearly threefold risk of obesity in those with both high genetic risk and with highest fast-food outlet exposure, relative to those at low genetic risk and least exposed to fast-food outlets.

The mechanisms by which BMI-associated SNPs confer a greater risk of obesity are not fully elucidated, but two SNPs included in our genetic risk score for BMI, FTO and MC4R, have been linked to dysregulation of appetite and loss of control over eating [29‐32], and preferences towards energy-dense [33], high-fat foods [23, 34]. As such, studies using genetic risk scores similar to the one used here have found that the association between these scores and obesity can be partly explained by appetite and eating behaviours [26, 35]. However, previous research has been limited to examining the interaction of BMI-associated SNPs with behavioural factors (gene-lifestyle interactions) such as diet and physical activity. For example, the influence of FTO on obesity risk can be moderated by physical activity [27]. Similarly, dietary risk factors for obesity, including consumption of sugar-sweetened beverages [36], and fried food [37], have been more strongly associated with weight gain in those with higher genetic obesity predisposition.

To our knowledge, this is the first published investigation of the interplay between any characteristic of the neighbourhood environment and genetic obesity risk. Our study was motivated to explore the hypothesis that genetic obesity risk would be exaggerated in permissive neighbourhood environments, characterised by an abundance of fast-food outlets, where large portions of affordable, energy-dense foods are readily accessible [38]. We expected that the strength of the relationship between fast-food outlet density and body weight would be stronger for those more genetically susceptible to obesity (Fig. 2a). We found no evidence to support this hypothesis. Alternatively, if it were the case that those with higher genetic risk were equally susceptible to the food environment, we would have expected the relationship between fast-food outlet density and body weight to be similar in both groups, albeit with those at higher risk having higher body weight at any given level of environmental exposure (Fig. 2b). However, although we found no evidence of interaction, our results indicated a possible third alternative, of a weaker association of the food environment with BMI among those with higher genetic risk of obesity, compared with those with lower genetic risk (Fig. 2c). We observed that, at highest levels of fast-food outlet exposure, mean BMI for the lower and higher genetic risk groups converged.

The key implication of our findings is the necessity for further research to test our observations and understand mechanisms in other samples. It is possible, for example, that the contribution of the food environment may simply be less detectable among a population of adults with higher risk of obesity stemming from multiple biological and metabolic determinants [25]. This explanation would be consistent with Rothman’s component cause model, which suggests that for outcomes with multiple causes, the apparent strength of any single causal component is influenced by the relative prevalence of other component causes [39]. Nonetheless, in the immediate term and over and above the contribution of genetic susceptibility, neighbourhood fast-food outlet exposure appears to be a significant and importantly modifiable contributor, subject to change via urban planning (‘zoning’) as a structural intervention to prevent obesity and promote the public’s health [40].

The major strengths of this study include objective assessment of neighbourhood fast-food outlet exposure, using accurate and contemporaneous food outlet location data; objective height and weight data, measured by trained researchers; and use of an empirically-derived genetic risk score based on 96 BMI-associated SNPs. The main limitation of this study is the cross-sectional, observational study design, which limits causal inference. It is possible that preference for fast food may drive the selection of residential neighbourhoods with abundant fast-food outlet access (reverse causality). Our estimates of combined associations might be confounded because preference for energy-dense fast food may arise as a function of genetic obesity predisposition. Our fast-food outlet exposure models may have residual confounding, for example from unobserved environmental attributes, such as exposure to food outlets beyond the home neighbourhood. This said, our models were comprehensively adjusted for behavioural and sociodemographic covariates.

Some misclassification may have occurred as our exposure (number of fast-food outlets) and outcome (adiposity) were measured at different time points (2011 and 2005–2015, respectively). This risk was minimised as far as possible through exposure data being collected at approximately the mid-point of the period of outcome data collection. Moreover, we minimised misclassification through operationalising fast-food outlet exposure in quartiles, which would have been less sensitive to food environment change over time. To test this, we cross-classified all participants, comparing their exposure classification based on 2011 food environment data, against that based on food environment data from 2014 (courtesy of the Food environment assessment tool, powered by Ordnance Survey) [41]. The resulting cross-classification was high (r_s = 0.94, P < 0.001). Therefore, any individual recruited into the Fenland Study between 2011 and 2014 would likely have been characterised similarly in terms of their fast-food outlet exposure. We have described the limitations of our neighbourhood fast-food outlet exposure metric previously [19, 22, 24].

We used a genetic risk score composed of 96 BMI-increasing alleles, which enabled us to detect a large main effect both independently and in combination with neighbourhood fast-food outlet exposure. However, this approach masks the contribution of individual SNPs, which may act through discrete behavioural or metabolic pathways of relevance to the food environment. Moreover, given the highly polygenic nature of body weight, genetic risk scores that use a genome-wide set of common variants in their construction may have stronger predictive power for obesity and BMI and should be considered for future research [42]. The Fenland Study is a population-based cohort study, with working-age adults recruited who were largely educated, employed and white British, living in Cambridgeshire, which is a county in the East of England comprising urban, suburban and rural areas, including the major cities of Cambridge and Peterborough. Characteristics of this population and study area, although common elsewhere in the UK and beyond, may influence the generalisability of our findings.

Our study confirms previously-identified associations of both genetic and neighbourhood risk factors with adiposity, and for the first time demonstrated independence in these relationships. These important drivers of the obesity epidemic have to date been studied in isolation. Although obesity-related SNPs contribute substantially to obesity in the population, neighbourhood fast-food outlet exposure also appears to be a significant determinant and an important and modifiable target of policy intervention to prevent obesity and promote the public’s health. We found no evidence of interaction between these determinants.