The Understanding Society study began in 2009, and uses surveys to learn about the lives of people from around 30,000 households across the UK. In 2010-2012, registered nurses visited around 20,000 respondents and asked them questions about their health. They also collected “biomarkers” including lung function, grip strength and weight, and for around 13,000 participants, a blood sample.
What makes Understanding Society special?
- Its representativeness: unlike clinical research cohorts, our participants are selected in a process that ensures the UK population is well-represented (in terms of people’s circumstances, ages and other attributes)
- Its focus on the household: this enables research on family dynamics
- The broad age range of respondents: this allows researchers to focus on different stages of the life course, such as old age
So, why should a social survey devote resources to a biomarker collection? To showcase the insights provided by these data, Understanding Society hosted an event on 30 May 2022 as part of a series celebrating its first ten years. Researchers who have used our biological data presented their findings across a range of topics, all of which explore the ways in which our environment interacts with our biology.
DNA methylation as a biomarker of smoking
Alex Andrayas (University of Essex) started the morning by presenting her PhD research, in which she used Understanding Society blood data to develop a biomarker of smoking. An epigenetic biomarker of smoking uses chemical changes to our DNA that are caused by smoking, which can be measured to estimate how much a person has smoked. Smoking may be underestimated by self-reports, and this might be different in different groups, because of ‘reporting bias’ (misleading or inaccurate responses), but epigenetic biomarkers can offer a more objective measure of smoking. This is important when trying to identify its role in health outcomes.
In her research, using data from Understanding Society and the 1958 National Child Development Study (NCDS), Alex found that people’s self-reported smoking status was more likely to agree with their smoking status assessed using the biomarker if they had fewer qualifications, or were in routine-to-intermediate occupations, compared to those with more qualifications, or with managerial-or-professional occupations. This suggests that smokers in a higher socioeconomic position might be more likely to under-report how much they smoke.
Alex then went on to share her research into the links between socioeconomic position, smoking and inflammation (a biological process thought to underlie many diseases). Both smoking and inflammation are known to be ‘socially patterned’ – i.e. they affect some socioeconomic groups more than others (for example, by age and by gender). Alex found that epigenetic biomarkers of smoking were more strongly associated with inflammation than self-reported smoking status was, and that inflammation differed by educational attainment and socioeconomic position after accounting for smoking using the biomarker. Smoking explained social differences in inflammation differently whether it was measured by self report or by the biomarker.
But what about passive smokers and former smokers? During the Q&A session, Alex explained that, in her research, it had proved challenging to disentangle the effects of passive smoking because smokers tended to live with other smokers, and that there is disagreement among researchers as to whether quitting smoking can fully reverse the epigenetic changes it causes. More research is needed to explore this further.
Does testosterone affect men’s earnings?
In contrast to most of the presentations, which focussed on how our environments might alter our biomarkers, Dr Amanda Hughes of the University of Bristol provided an example of how what’s ‘under the skin’ might shape our experiences when she presented her research on testosterone, risk and socioeconomic position in British men.
Testosterone is a steroid hormone that is present at much higher levels in men compared to women, and is important in the development of male bodies. Its role in behaviour has also been studied, including risk-taking behaviours; after the 2008 economic crash there was an increase in media interest regarding the role of testosterone in high-risk financial sector investments, which precipitated several pieces of research into testosterone and a range of economic outcomes. The direction of effect is not always clear, as testosterone level can be affected by our health, and by events in our lives which affect our self-esteem.
Exploring cause and effect
When a research question is hindered by possible confounders and reverse causation, causal relationships can be estimated using a technique called Mendelian Randomization. This uses common variations in our DNA that are associated with a possible ‘exposure’ (such as increased testosterone in the blood) to see whether they are robustly associated with a given outcome (such as earnings). This approach assumes that we can estimate people’s tendency toward a higher or lower level of testosterone using a ‘polygenic score’ for testosterone, which is based on how many copies of testosterone-raising genetic variants each person has. While testosterone level is affected by many non-genetic factors, people with a higher polygenic score should – on average – have more testosterone in their blood than people with a lower polygenic score.
Amanda used this approach to estimate the effect of testosterone on a range of occupational, educational and risk-taking outcomes using Understanding Society data, and compared these results to those obtained when she used actual measures of testosterone in the blood. She found that two of the eight outcomes – gross earnings and the probability of being in work – were positively associated with the testosterone polygenic score, but not with actual testosterone. She also compared testosterone levels in men with different income and education levels (high, middle or low) and found no difference. This is a new finding and contradicts earlier research – Amanda suggested that this difference may stem from the fact that Understanding Society is more representative of the general population. Amanda concluded by saying that she would like to follow up her work with longitudinal analyses, i.e. using a second testosterone measurement for the same set of participants, to explore changes over time.
Could indicators of stress influence housing policy?
Dr Amy Clair, from the University of Adelaide, spoke on how biomarkers can improve our understanding of the well-established link between housing and health. With modern improvements in housing conditions, attention has increasingly shifted from “hard” components of housing such as sanitation and heating, to “soft” components such as the stress of insecure tenure leading to anxiety and depression. Biomarkers can tell us whether study participants exhibit risk factors for future ill health, and can reveal undiagnosed illnesses.
Interestingly, both physical and psychological challenges can be reflected in the level of a protein in our blood called CRP, increased levels of which can indicate stress, infection and/or inflammation. In her research, Amy used Understanding Society data to measure CRP in 9,593 individuals aged between 22 and 95. Individuals with the highest levels of CRP – indicating a recent infection – were excluded, meaning that any findings should reflect chronic (long-term) physiological processes. Associations between CRP and housing were analysed using the wide range of housing measures afforded by Understanding Society’s rich survey data. CRP was higher in private renters, and lower in people living in detached houses, indicating a link between our housing situation and our wellbeing.
Amy then discussed the policy implications of her findings: as the health and wellbeing of private renters appears to be worse, policies should aim to improve the private rented sector as well as increase social housing, reducing dependence on the private rented sector, e.g. by ending the Right to Buy scheme (forty per cent of homes bought under this scheme were privately rented in 2017).
During the Q&A, Professor Meena Kumari suggested that it might be interesting to study the individuals with CRP levels that indicate recent infection, rather than exclude them, as infection can result from poor housing conditions such as cold and damp. Neighbourhood-level factors were also highlighted as a good area for future research.
The link between job quality and physiological health
In the final presentation of the day, Professor Tarani Chandola, from the University of Hong Kong, highlighted the usefulness of biomarker data compared to questionnaire data and spoke about his research on re-employment, job quality and allostatic load. ‘Allostatic load’ refers to a physiological state, brought on by chronic stress, in which several bodily systems become dysregulated. It is linked to physical illness and is calculated in individuals by measuring a range of biomarkers and counting how many of them are above (or below) an ‘at-risk’ cut off point. Previous research has shown that returning to work after a period of unemployment improves physical and mental health, but that this depends on job quality and security.
Is ‘bad work’ better than ‘no work’?
Tarani made use of the biomarkers and detailed job quality data in Understanding Society to explore whether ‘bad work’ was better for people than unemployment, using allostatic load to measure wellbeing. He focussed on individuals aged 30 to 75 who were unemployed during their Wave 1 interview. Blood samples and current job information were collected across Waves 2 and 3. Job quality indicators were low pay, job insecurity, low job control, job dissatisfaction and job anxiety. Respondents were assigned to one of four categories: remained unemployed, re-employed into good quality work, re-employed with one of the aforementioned low quality indicators, and re-employed with two or more low quality indicators. To measure wellbeing, an allostatic load index was calculated by adding the number of biomarkers for which an individual had ‘at-risk’ measures. Income and self-reported mental and physical health were also measured as outcomes. While household income improved for those who were re-employed compared to those who weren’t, wellbeing varied between groups. Self-reported mental health was best among those re-employed into a good quality job, and was worst among those who remained unemployed. However, the group with the worst (highest) allostatic load was the re-employed group with the worst quality jobs. These results took into account a number of factors including social factors and the health of those that ended up with poor quality jobs. During the Q&A, Tarani noted that his findings are perceived differently by UK and US audiences, possibly due to cultural and welfare system differences.
Interested in using our biological data?
- We have measurements of 21 health-related biomolecules (biomarkers), two DNA-derived datasets (genotypes and methylation) and proteomics (measurements of 184 proteins)
- Our biomarker topic webpage gives an overview of how biomarkers can be used in social research
- Professor Meena Kumari has recorded a video describing the nurse visit and biological data
- Research using Understanding Society’s biological data can be found in our Publications library
- Find out how to access the data or contact Meena
Authors
Anna Dearman
Anna is a PhD student at the Institute for Social and Economic Research at the UNiversity of Essex, in the Soc-B Centre for Doctoral Training
