Introduction
Annually there are approximately 17 million cardiovascular disease deaths worldwide (WHO, 2002), with high blood pressure contributing to this mortality (Kearney et al., Reference Kearney, Whelton, Reynolds, Muntner, Whelton and He2005). Factors resulting in high risk for cardiovascular diseases have been well studied in high income countries, but have been less well researched in low/middle income countries. This is despite robust theoretical evidence of environmental transitions in these countries which could result in biological adaptations that increase cardiovascular disease risk (Adair & Prentice, Reference Adair and Prentice2004). Given the different social, economic and cultural contexts of transitioning societies compared with high income countries, it is important to better understand how such environments influence cardiovascular disease risk.
This research uses the principles of embodiment, which recognize that humans are both social and biological beings. Embodiment refers to how human beings ‘incorporate, biologically, the world in which we live, including our societal and ecological circumstances,’ (Krieger, Reference Krieger2005, p. 351). In Glass & McAtee's (Reference Glass and McAtee2006, p. 1653) model, which fosters these principles of embodiment to study behaviour and disease, time from conception to death runs along a horizontal axis representing the life course. In the model a vertical axis exists that incorporates the different levels of social organization that influence health (e.g. family, school, neighbourhood, state, geopolitical systems) as well as the biological levels that embody the impact of the social environments (e.g. multi-organ system, cellular, molecular and genomic substrate levels). The model helps to show that health outcomes are influenced over time by human actions and behaviours as individuals move through different social and physical environments and embody these at varying biological levels at different stages of the life course (Glass & McAtee, Reference Glass and McAtee2006). This paper will particularly focus on considering the SES environment at two levels (household and neighbourhood) and at two points in the life course (infancy and age 16 years). The paper uses data from Birth to Twenty (Bt20), a 1990-born Johannesburg–Soweto South African birth cohort, to study associations between the socioeconomic environment and blood pressure outcomes in adolescents.
The South African context provides a unique transitioning environment to study associations between socioeconomic status (SES) and blood pressure. The Bt20 cohort was born in the year that Nelson Mandela was released from prison. The four decades preceding this had seen the introduction of legislation endorsing apartheid in South Africa. Apartheid legislation split the population into four ‘population groups’ that were labelled African or Black, Mixed race or Coloured, Asian or Indian, and White. The combined impact of apartheid and the economic sanctions placed on the South African government because of apartheid legislation by the international community resulted in vast inequality between the population groups by the mid-1980s (Cameron, Reference Cameron2003). Health expenditure for Whites in the mid-1980s by the government was five times that spent on the Black population per capita, and the infant mortality rate was also five times higher in the Black group. At the same time literacy was less than 50% in the Black population compared with 99% in the White group (Cameron, Reference Cameron2003). Since 1990 South Africa's Human Development Index has remained quite stable changing from 0.601 in 1990 to 0.597 in 2010, resulting in a drop in seven places in global rankings but maintaining a classification in the medium development level of the global table. South Africa is one of the more developed countries in sub-Saharan Africa, with only Botswana, Namibia and Gabon having higher rankings on the Human Development Index in 2010 (UNDP, 2011). Risk for chronic disease in South Africa is higher than in most other sub-Saharan African countries because of its relatively higher economic development, which has fuelled its more rapid transition. For instance WHO (2011) data show that it ranks second in sub-Saharan Africa for its male (6th in Africa) and female (3rd in Africa) adult (aged 15+ years) prevalence of overweight (68.5% females and 41.3% males with a BMI>25) (WHO, 2011). South Africa therefore represents a middle income nation that has been undergoing significant political and social change with a high prevalence of risk factors for cardiovascular disease such as overweight. Understanding the associations between SES and risk for cardiovascular disease within this context will build upon the already established evidence base relating to these associations in high income settings.
The evidence base from high income settings shows that socioeconomic status is known to be associated with adult cardiovascular disease and high blood pressure. A review of adult studies found that lower SES was associated with higher blood pressure in most high income country studies, although consistent associations were not observed in low/middle income countries (Colhoun et al., Reference Colhoun, Hemingway and Poulter1998). The majority of childhood/adolescent studies, including most low/middle income country studies that were published after the review of Colhoun et al. (Reference Colhoun, Hemingway and Poulter1998), have also not found SES-associated differences in blood pressure (Miranda et al., Reference Miranda, Davies, Smith, Smeeth, Cabrera and Gilman2009; Adair et al., Reference Adair, Martorell, Stein, Hallal, Sachdev and Prabhakaran2009). Only two published studies in low/middle income countries have shown an association in children/adolescents, such that low SES is correlated with increased blood pressure (Akinkugbe et al., Reference Akinkugbe, Akinwolere and Kayode1999; Longo-Mbenza et al., Reference Longo-Mbenza, Lukoki Luila and M'Buyamba-Kabangu2007). Longo-Mbenza et al. (Reference Longo-Mbenza, Lukoki Luila and M'Buyamba-Kabangu2007), however, failed to take account of height, a critically important variable in analyses of blood pressure in children/adolescents, and Akinkugbe et al. (Reference Akinkugbe, Akinwolere and Kayode1999) did not use appropriate age- and sex-adjusted standards for assessing blood pressure (US Department of Health and Human Services & National Institutes of Health Heart Blood and Lung Institute, 2005). Focusing on adolescent cardiovascular risk factors in low/middle income country contexts is important in order to understand critical life-course associations between SES and blood pressure because there is some suggestion from high income countries that inequalities might equalize during the adolescent period (West, Reference West1997).
Biological changes in metabolic, endocrine and/or immune parameters early in life due to poor environmental conditions have been proposed to be responsible for links between the environment and later health outcomes (Lucas, Reference Lucas, Bock and Whelan1991). Biologists often refer to these physiological changes as adaptations (Ellison, Reference Ellison2005; Kuzawa, Reference Kuzawa2005). Potential adaptations that lead to increased blood pressure include alterations in kidney structure and function, epigenetic mechanisms such as those that might be induced by short telomeres, endocrine mechanisms such as IGF-1's role in the growth of vascular smooth muscle cells and variation in body composition (Adair & Dahly, Reference Adair and Dahly2005). It has been proposed that early-life adaptations may be based on predictions that arise not only from the mother's own environmental experiences but also those of preceding generations (Kuzawa, Reference Kuzawa2005). For the Bt20 participants the end of apartheid has meant that the cohort have been born into a rapidly changing social environment. If adaptations such as the ones proposed by Kuzawa do occur they could be more problematic in the type of environment that the Bt20 cohort were born into, because of the rapidly changing environment. This could result in predictions based on the maternal environment being inappropriate for the cohort participants' growth and development (Kuzawa, Reference Kuzawa2005).
In relation to blood pressure, there is a clear need for research to consider the potentially changing role of SES in its association with blood pressure across ages (Colhoun, Reference Colhoun, Hemingway and Poulter1998). A number of studies have taken a life-course approach to studying health effects of SES; in high income countries a literature review identified gaps in the literature to be a lack of prospectively collected childhood SES data collected at key developmental stages or a lack of SES data collected on more than two occasions, and a lack of studies reporting results for non-Whites (Pollitt et al., Reference Pollitt, Rose and Kaufman2005).
The association between blood pressure and individual/household SES has been more researched than neighbourhood SES. Households with similar SES can have different health outcomes when living in contrasting areas (Macintyre & Ellaway, Reference Macintyre, Ellaway, Berkman and Kawachi2000). In reviews, Pickett & Pearl (Reference Pickett and Pearl2001) and Riva et al., (Reference Riva, Gauvin and Barnett2007) found that contextual (i.e. neighbourhood) effects existed in most studies. The majority of recent evidence from high income settings highlights significant associations between deprivation in neighbourhoods and risk factors for coronary artery disease such as high blood pressure (Diez-Roux et al., Reference Diez-Roux, Kershaw and Lisabeth2008; Mujahid et al., Reference Mujahid, Diez-Roux, Morenoff, Raghunathan, Cooper, Ni and Shea2008). Neighbourhood SES could be linked to blood pressure through its known influence on health behaviours like physical activity or availability of healthy foods (Duncan et al., Reference Duncan, Spence and Mummery2005). It could also influence blood pressure through psycho-social pathways as poor environments could result in elevated psychosocial stress leading to an abnormal neuroendocrine secretory pattern (Galea et al., Reference Galea, Ahern, Rudenstine, Wallace and Vlahov2005). For the Bt20 cohort apartheid legislation limited the neighbourhoods that population groups could reside in around the time of the cohort's birth, although the removal of this legislation very quickly led to children experiencing different environments. For example between birth and 14 years, 56% of the Black cohort children had moved their address at least once (Ginsburg et al., Reference Ginsburg, Norris, Richter and Coplan2009). This means that by the time the children of the cohort were entering their adolescent years they were residing in very diverse neighbourhoods, with the potential to influence health in different ways.
Using Glass & McAtee's (Reference Glass and McAtee2006) model, two gaps in the literature have been identified in relation to understanding the association between SES and blood pressure outcomes in adolescents from low/middle income countries: 1) the lack of use of a life-course perspective to assess inequalities (the horizontal axis of Glass & McAtee's model); and 2) a focus only on household inequalities without placing them within the wider social context (the vertical axis of Glass & McAtee's model). Thus the present paper aims to study associations between household SES in infancy, household SES at age 16 years, neighbourhood SES at 16 years and systolic blood pressure, as well as systolic pre-hypertension, in a sample of Black 16-year-old urban South Africans.
Methods
Participants
The participants in this study were participants in the Bone Health sub-cohort of the Bt20 study, which has been described in detail elsewhere (Richter et al., Reference Richter, Norris and De Wet2004, Reference Richter, Norris, Pettifor, Yach and Cameron2007). At ages 9/10 years, a random sub-sample from the cohort (n=429) were enrolled into a longitudinal study assessing factors influencing bone health. Bone Health participants had more detailed health assessments than the Bt20 cohort. The Bone Health participants were predominantly Black participants (n=376) as a consequence of the demographics of the South African population. Only Black participants are included in this analysis because of the small sample size of the White ethnic group and its homogeneity within the high-SES group. Those Black participants with data on weight, height and systolic blood pressure that were measured at 16 years of age were included in current analyses (n=358, 47% female).
Ethical approval was granted by the ethics committees of the University of the Witwatersrand, South Africa, for primary data collection, and Loughborough University, UK, for secondary data analyses. The primary caregiver gave written informed consent for their adolescent to participate in each assessment and the adolescent provided written assent at 16 years.
Socioeconomic status measures
During infancy and at 16 years, caregivers were asked to assess household SES using a questionnaire based on standard measures used by the Demographic and Health Surveys (see www.measuredhs.com). The Bt20 SES questionnaire was piloted with 30 non-cohort caregivers to ensure understanding of concepts, an optimal layout and to test translation. Measures included caregiver's education, private medical insurance coverage, home ownership, housing type, water/toilet facilities, marital status and consumer durable ownership.
At 16 years of age, neighbourhood SES was assessed using a culturally relevant questionnaire, which was specifically developed in the Bt20 cohort using focus group discussions and in-depth interviews (Sheppard et al., Reference Sheppard, Norris, Pettifor, Cameron and Griffiths2010). Questions related to economic and social aspects of neighbourhoods, as well as the school environment, were included. The adolescents answered the neighbourhood/school SES questions. The concept of a ‘sociologically meaningful’ (Bond Huie, Reference Bond Huie2001) definition of neighbourhood was used, and the findings of an earlier qualitative study were adopted to define a neighbourhood covering an area within approximately 20 minutes walk (i.e. 2 km) of home in any direction (Sheppard et al., Reference Sheppard, Norris, Pettifor, Cameron and Griffiths2010). Convenient administrative boundaries were not used to define neighbourhoods as others have suggested that such boundaries do not necessarily associate with actual areas defining the causal variables linking neighbourhood social environments to health (Pickett & Pearl, Reference Pickett and Pearl2001).
Anthropometric measures
Birth weight, and weight and height at 16 years, were assessed using standard techniques (Lohman et al., Reference Lohman, Roche and Martorell1991). Low birth weight was defined as a birth weight less than 2.5 kg. Body mass index (BMI) was calculated as weight (kg)/height (m)2. Adolescents were classified as normal weight, overweight or obese using Cole et al.'s (Reference Cole, Bellizzi, Flegal and Dietz2000) international age-specific cut-off points.
Blood pressure measures
Blood pressure was measured three times using a digital device (Omron M6; Omron, Kyoto, Japan) at 16 years of age. Appropriate cuff sizes were used, and participants were measured seated and resting with a break of several minutes between measurements. The first measurement was discarded and the second and third averaged. An average systolic blood pressure at or above the National Heart Blood Lung Institute's (US Department of Health and Human Services & National Institutes of Health Heart Blood and Lung Institute, 2005) 90th sex-, age- and height-standardized percentile was used to define systolic pre-hypertension. Systolic pre-hypertension, and not stage 1 hypertension, was included in the current analyses because few adolescents in this sample had stage 1 hypertension.
Other variables used in the analyses
Individuals born before 37 weeks gestation were classified as pre-term and after 41 weeks as post-term. Mother's marital status and age were self-reported during infancy. Adolescent's parity and smoking status (current, previous or never) were self-reported in a questionnaire.
Statistical analyses
Linear regression was used where the continuous systolic blood pressure variable was the outcome and logistic regression for the dichotomous systolic pre-hypertension outcome. Because systolic blood pressure is known to vary by sex and height, all analyses where systolic blood pressure was the outcome were stratified by sex and adjusted for height. Where data were missing for a particular variable they were coded into a separate category to maximize the sample size for analysis. Using this approach, with alpha set at 0.05 and power at 0.80, the sample size analysed has the statistical power to detect medium to large effect sizes, but would not be adequate to detect small effect sizes (Cohen, Reference Cohen1992). Initial unadjusted regression analyses explored relationships between each SES measure and each blood pressure outcome (controlling for height for systolic blood pressure). Subsequently multivariate regression analyses were adjusted for all variables that had shown a relationship with systolic blood pressure or systolic pre-hypertension in the initial analyses with a p-value <0.1.
Regression models were built in steps: 1) infancy SES variables from the initial analysis (p<0.1) entered; 2) added year-16 household/neighbourhood SES variables with a p-value <0.1 from the initial analysis; and 3) added other infancy and year-16 variables with a p-value <0.1 in unadjusted analyses. This approach allows for any mediating effect of the year-16 SES variables on the association between the infancy SES variables and the systolic blood pressure outcomes, as well as any mediating effects of the other infancy and 16-year measures on the association between the SES measures and systolic blood pressure to be investigated.
Over 100 questions were asked relating to neighbourhood SES. To enable a more parsimonious analysis of these measures and to avoid potential problems of multicolinearity, principal components analysis (PCA) was used to extract indices that grouped similar neighbourhood SES variables together. A theory-based approach was used to develop nine neighbourhood indices and PCA confirmed the appropriateness of grouping these variables together. In each case the first component scores were extracted and the statistical assumption that all eigenvalues be greater than 1 was met. Standardized Cronbach alpha scores were calculated as a measure of reliability and the percentage of the variation in the variables explained by the PCA index calculated. Three indices measured neighbourhood economics: 1) economic index, 2) need for more services/facilities index, and 3) problem index. Two indices measured neighbourhood social aspects: 1) crime prevention index and 2) social support/happiness index. In addition there were two variables (‘How safe do you feel in the neighbourhood’ and ‘How much crime is there in the neighbourhood?’) that did not load well onto any indices and were therefore retained as individual variables. There were also two school neighbourhood indices identified: 1) school environment index and 2) school problems index. In addition to the seven neighbourhood SES indices, household SES questionnaire data were used to construct two indices that measured ownership of consumer durables, the first during infancy and the second at 16 years. Regression factor scores were extracted for each index and tertiles were created. The highest tertile or group represented the highest level of SES (reference category).
All analyses were conducted using SPSS 16.0 (Chicago, Illinois).
Results
Descriptive statistics
The prevalence of systolic pre-hypertension at age 16 years in this sample of Black South African adolescents was approximately 11%, with a mean (SD) for systolic blood pressure of 110.4 mmHg (9.4) and 116.3 mmHg (10.4) for girls and boys respectively (Table 1). Boys were significantly heavier than girls at birth (two-sided p<0.05), but at 16 years of age girls were approximately three times more likely to be overweight/obese compared with boys (χ2=24.02, p<0.001).
Table 1. Mean (SD) systolic blood pressure, birth weight, height and body mass index (BMI) and percentage with pre-hypertension, stage 1 hypertension, overweight and obese by sex for Black South African adolescents aged 16 years
a Pre-hypertension is defined using a systolic blood pressure at or above the National Heart Blood Lung Institute's 90th percentile sex-, age- and height-standardized value (US Department of Health and Human Services & National Institutes of Health Heart Blood and Lung Institute, 2005). Stage 1 hypertension is defined using a systolic blood pressure at or above the National Heart Blood Lung Institute's 95th percentile sex-, age- and height-standardized value plus 5 mmHg.
b Ten cases did not have birth weight recorded.
c Overweight and obesity are defined using Cole et al.'s (Reference Cole, Bellizzi, Flegal and Dietz2000) age-appropriate international cut-offs for children and adolescents.
* Indicates a significant (p<0.05) sex difference in this variable. Continuous variables were tested using an independent samples t-test and categorical variables using a multidimensional Chi-squared test.
Household and neighbourhood SES indices
The distribution of the tertiles of household assets at birth and 16 years shows the dynamic nature of relative SES between birth and age 16 years in the cohort (Fig. 1). For example, of the households classified as low SES at birth using this asset index, just over 50% were classified as medium or high at age 16 years.
Fig. 1. Transitioning relative household socioeconomic profile of the sample between birth and age 16 years.
The constructs included in each of the neighbourhood indices are shown in Table 2. Cronbach alpha values for the indices varied from 0.56 to 0.98, with between 27 and 91% of the variation explained by the various PCA indices. These values are higher than those observed for neighbourhood indices reported with UK data (Cronbach's alpha scores between 0.48 and 0.98, with two scores lower than this range considered to be too low for internal reliability in the UK study) (Cummins et al., Reference Cummins, Macintyre, Davidson and Ellaway2005).
Table 2. Neighbourhood principal component analysis (PCA) indices and percentage of variation explained by the extracted index
Systolic blood pressure
Initial linear regression models of systolic blood pressure at 16 years controlling for height showed that for males infancy variables that resulted in significantly higher systolic blood pressure (other than the missing data category) were being born post-term, living in a private rental property compared with owning, and having a mixture of inside/outside or outside-only water/toilet facilities (Table 3). Birth weight was significantly negatively associated with systolic blood pressure for boys. For females the only one of these infancy variables that was shown to be significantly associated with increased systolic blood pressure was being born post-term. In addition, for females having a mother who was aged 20–29 years at the time of birth was associated with significantly reduced systolic blood pressure (Table 3). At 16 years for males, living in a neighbourhood with not much crime relative to a lot of crime was associated with lower systolic blood pressure, and those in the middle tertile of the neighbourhood problem index had significantly higher systolic blood pressure than those in the highest tertile (Table 4). Furthermore, for males being taller and having a higher BMI or being obese were also associated with higher systolic blood pressure. When compared with males, the females had a broader range of SES measures that were found to be significantly associated with systolic blood pressure adjusting only for height in initial regression models at age 16 years. Being in the middle tertile of the neighbourhood need for more services index was associated with having lower systolic blood pressure compared with those in the highest tertile. Those in the lowest tertile of the neighbourhood crime index, school environment index, the school problems index or the household consumer durable index were predicted to have increased systolic blood pressure compared with those in the highest tertile of these indices (Table 4). Similar to the male findings, females with a higher BMI or who were obese were associated with higher systolic blood pressure, although unlike the male findings being taller was not significantly associated with increased systolic blood pressure in the females (Table 4).
Table 3. Infancy predictors of systolic blood pressure from initial regression analysesa for Black males and females and unadjusted odds ratios for pre-hypertension (n=358)
a Adjusted for height at age 16 years.
b Ref.=reference category.
*p<0.1; **p<0.05; ***p<0.01; ****p<0.001 (two tailed).
Table 4. Year-16 predictors of systolic blood pressure from initial regression analysesa for Black males and females and unadjusted odds ratios for pre-hypertension (n=358)
a Adjusted for height at age 16 years.
b Ref.=reference category.
c Overweight and obesity are defined using Cole et al.'s (Reference Cole, Bellizzi, Flegal and Dietz2000) age-appropriate international cut-offs for children and adolescents.
*p<0.1; **p<0.05; ***p<0.01; ****p<0.001 (two tailed).
For males, none of the infancy or year-16 SES variables was significant in its association with systolic blood pressure after being entered simultaneously into regression models (Table 5). Lower birth weight and increased height were significantly associated with increased systolic blood pressure in the adjusted model, which accounted for 13.1% of the variance in systolic blood pressure for males (Table 5). Neither of these variables was significant in the female models of systolic blood pressure. For females the only significant variables in the fully adjusted models showed that those in the middle tertile of the index of crime prevention showed significantly higher systolic blood pressure compared with those in the highest tertile. Furthermore, obese girls and those born post-term were also associated with significantly higher systolic blood pressure. The fully adjusted model for females explained 23.2% of the variation in systolic blood pressure (Table 5).
Table 5. Adjusted unstandardized parameter estimates for systolic blood pressure in Black 16 year olds for variables that had a previous significant association with systolic blood pressure in models adjusting for height only
a Multivariable regression models were adjusted for year-16 height and were built in steps: 1) infancy variables with a p-value <0.1 from the initial analysis entered; 2) added year-16 household/neighbourhood SES variables with a p-value <0.1 from the initial analysis; and 3) added other year-16 variables with a p-value <0.1 from the initial analysis.
b Ref.=reference category.
c Overweight and obesity are defined using Cole et al.'s (Reference Cole, Bellizzi, Flegal and Dietz2000) age-appropriate international cut-offs for children and adolescents.
d This variable violates the colinearity assumptions for the female model and could therefore not be included in the multivariate model.
*p<0.05; **p<0.01; ***p<0.001 (two tailed).
Systolic pre-hypertension
The only infancy variable that was significantly associated with reduced risk of systolic pre-hypertension in the initial model (other than the missing data category) was maternal age 25–29 years compared with a maternal age of 35 plus years. Increased risk of systolic pre-hypertension was observed in those who were born post-term, and who had outside-only access to water/toilet facilities and shared use of water and toilet facilities (Table 3). In initial models for year-16 SES variables, shared use of indoor running cold water or shared use of an outdoor flush toilet, compared with sole use, were significantly associated with increased risk of systolic pre-hypertension (Table 4). In step one of the multiple regression, all of the significant infancy variables from the initial models, apart from shared use of water and toilet facilities, retained significance and direction of association (Table 6). In step two, the only infancy variable to retain significance was having a mother aged 25–29 years. None of the year-16 SES variables was significant when entered simultaneously in step two. Step three did not change the significance of the other variables, although those born post-term were also found to have higher systolic pre-hypertension risk than those born at term.
Table 6. Adjusted odds ratios for risk of systolic pre-hypertension in Black 16 year olds for variables that had a significant unadjusted odds of pre-hypertension
a Multivariable regression models were built in steps: 1) infancy variables with a p-value <0.1 from the initial analysis entered; 2) added year-16 household/neighbourhood SES variables with a p-value <0.1 from the initial analysis; and 3) added other year-16 variables with a p-value <0.1 from the initial analysis.
b Ref.=reference category.
c Overweight and obesity are defined using Cole et al.'s (Reference Cole, Bellizzi, Flegal and Dietz2000) age-appropriate international cut-offs for children and adolescents.
*p<0.05; **p<0.01; ***p<0.001 (two tailed).
Discussion
Within the context of Glass & McAtee's (Reference Glass and McAtee2006) framework this paper has provided evidence that enhances understanding of associations between two of their identified levels of hierarchy of SES measurement (household and neighbourhood) with systolic blood pressure measures in a middle income context. Furthermore, the paper has considered the stage of the life course at which the Bt20 household SES measures were assessed (infancy and age 16 years). Glass & McAtee also acknowledge the importance of considering the life course stage at which environmental differences are embodied in their model. These associations have been studied within the context of a cohort of South African adolescents born at the dawn of democracy in South Africa after a prolonged period of legislation-enforcing apartheid. Despite the legacies of apartheid, South Africa is one of the most advanced countries for economic, social and epidemiological transition in the sub-Saharan African region. Results are discussed within the context of the two levels of hierarchy studied (household and neighbourhood), as well as an overall consideration of the relative contribution of the different life-course measurement time points of birth and age 16 years for the systolic blood pressure outcomes considered.
Infancy SES measures
Despite the broad range of infancy-assessed household SES measures tested, none remained significant in adjusted analyses. This finding of a lack of association is not unique to this study (Lawlor et al., Reference Lawlor, Ebrahim and Davey Smith2002). Where an SES association was observed in the unadjusted analyses, it followed the pattern previously described in the literature of a negative association between SES measured in infancy and disease risk (Pollitt et al., Reference Pollitt, Rose and Kaufman2005).
Most of the significant infancy SES variables in unadjusted analyses relate to early-life measures of water and sanitation facilities. Whilst these measures are probably a proxy for an overall measure of the SES environment, it is also feasible that aspects of the SES environment specific to water and sanitation are more important for systolic blood pressure in early life than other SES measures. Poor sanitation facilities might result in a high risk environment for infections with a faecal–oral route of transmission (e.g. diarrhoea). High diarrhoea prevalence in infancy could result in repeated exposure to dehydration, which may ‘programme’ the body for increased water and salt retention to cope with a risky dehydration environment, leading to hypertension risk (Davey Smith et al., Reference Davey Smith, Leary and Ness2006; Lawlor et al., Reference Lawlor, Smith, Mitchell and Ebrahim2006). This hypothesis might be better tested in other countries in sub-Saharan Africa with higher prevalence of diarrhoea and malnutrition early in infancy because such countries would provide higher prevalence levels of repeated severe diarrhoea episodes. Most studies investigating this association between diarrhoea episodes/sanitation and blood pressure have taken place using high/middle income country cohorts in the UK, Brazil and Peru and have not provided evidence for this association (Batty et al., Reference Batty, Smith, Fall, Sayer, Dennison, Cooper and Gale2007, Reference Batty, Horta, Smith, Barros and Victora2009; Pearce et al., Reference Pearce, Relton, Unwin, Adamson and Smith2008; Miranda et al., Reference Miranda, Davies, Smith, Smeeth, Cabrera and Gilman2009). The fact that Bt20 data show some evidence in unadjusted analyses in a middle income country context to potentially support this hypothesis, suggests that further testing in low income environments with higher prevalence of diarrhoea is warranted.
Age 16 years household SES measures
Similar to the findings for the infancy SES variables, none of the year-16 SES measures was significant in adjusted analyses. In previous child/adolescent developing country studies in adjusted analyses where height, age and sex were accounted for, no association between SES and blood pressure was shown, which concurs with what has been observed here (Adair et al., Reference Adair, Martorell, Stein, Hallal, Sachdev and Prabhakaran2009; Miranda et al., Reference Miranda, Davies, Smith, Smeeth, Cabrera and Gilman2009).
There are not many other studies that have simultaneously assessed the association between household and neighbourhood SES measures and blood pressure in adolescents. Chen & Paterson (Reference Chen and Paterson2006) studied these associations in a US sample of 315 14–19 year olds and observed little association between measures of household SES and systolic blood pressure before controlling for neighbourhood SES measures. Only one measure out of four tested showed an inverse association with systolic blood pressure and this was family savings (R 2=0.02) (Chen & Paterson, Reference Chen and Paterson2006). This compares with an R 2=0.08 in the current study for females in a model that included only water and toilet facilities at age 16 years and height. Our findings are also similar to those in a sample of 212 US 14–16 year olds undertaken by McCrath et al. (Reference McGrath, Matthews and Brady2006), who considered the association between household education/occupation levels and systolic blood pressure and observed no significant effect. A limitation of comparing the findings of this study with both of these preceding studies is that neither of these papers controlled for height, which could alter their findings. Nevertheless, the combined results of these studies suggest that household SES measures do not show a strong association with systolic blood pressure in adolescents aged 14–19 years. Our results, along with the weak associations presented in previous adolescent studies, provide support for West's (Reference West1997) hypothesis of equalizing of health inequalities in adolescence.
Age 16 years neighbourhood SES measures
This study aimed to understand the relationship between neighbourhood SES and systolic blood pressure in a transitioning country to add to the evidence base already built in high income countries. In adult studies in high income countries the association between neighbourhood SES and cardiovascular disease risk factors has been observed in the majority of studies, with those in higher SES neighbourhoods having less risk for cardiovascular disease (Diez Roux et al., Reference Diez-Roux, Kershaw and Lisabeth2008). Findings of this study in 16-year-old children show that in adjusted analyses only one measure of neighbourhood SES was significant in its association with systolic blood pressure for females only. Those in the middle tertile of the index of crime prevention showed higher systolic blood pressure than those in the highest tertile. This association was, however, not significant for those in the low tertile of the index, although the association was in the same direction. The lack of association in the lower tertile could be related to the relatively smaller sample size in this group amongst females in the sample compared with the middle tertile, which would reduce the statistical power available to detect differences. Similar studies that have assessed the association between neighbourhood SES, household SES and systolic blood pressure in high income countries in adolescents have observed contrasting findings regarding the significance of associations. Chen & Paterson (Reference Chen and Paterson2006) revealed no significant association between systolic blood pressure and neighbourhood measures of SES (education levels, employment, income and housing values) in unadjusted analyses. McGrath et al. (Reference McGrath, Matthews and Brady2006) observed an association between a neighbourhood measure of income at or below the poverty line in both adjusted and unadjusted analyses with systolic blood pressure. No association was observed with other neighbourhood measures (racial composition and low levels of high school education). The differences in results across the adolescent studies of the association between household/neighbourhood SES and systolic blood pressure could be partly explained by the different measures of neighbourhood SES used, different ages within adolescence examined, and the different methods adopted to define neighbourhoods. Whilst all three studies had similar objectives, comparisons between the findings are complicated because of these differences. These problems are common to comparisons of findings across neighbourhood-level studies (Bond Huie, Reference Bond Huie2001; Messer, Reference Messer2007; Mujahid et al., Reference Mujahid, Diez-Roux, Morenoff and Raghunathan2007).
Most adult studies of SES and blood pressure have shown BMI to mediate the association (Colhoun et al., Reference Colhoun, Hemingway and Poulter1998). This mediation did not occur with the neighbourhood crime measure in this study. This may be because the significant associations observed with this SES measure could work through different mechanisms besides BMI. Although a lack of neighbourhood crime prevention could make healthy lifestyle behaviours (e.g. physical activity) more difficult, it is also plausible that the influence of such neighbourhoods on blood pressure could be mediated through stress, for example through increased cortisol secretion, which increases risk of higher blood pressure (Kapuku et al., Reference Kapuku, Treiber and Davis2002).
The contribution of the birth and age-16 environment to age-16 systolic blood pressure
When considering the time (life course) axis of Glass & McAtee's (Reference Glass and McAtee2006) framework, the infancy measures had a slightly higher overall contribution to the variation in systolic blood pressure outcomes for males (Adj. R 2=0.07) compared with the age-16 SES measures (Adj. R 2 change=0.02). For females, the contribution of the SES measures at age 16 to the proportion of variance explained in systolic blood pressure was greater (Adj. R 2 change=0.05) than the proportion of variation explained by the infancy SES measures (Adj. R 2=0.01), suggesting a sex difference in the relative importance of the SES measures taken at different ages. However, for both males and females the overall variation in systolic blood pressure explained by any of the SES measures, regardless of the point on the horizontal time axis at which they were measured, is small. This shows little evidence to support physiological changes related to adolescent systolic blood pressure having taken place as a result of environmental insults early in life in this middle income context, even during this time of social change in South Africa. However, it would be dangerous to make strong conclusions at this stage in relation to early-life influences on later risk for adult disease as this adolescent sample have relatively healthy systolic blood pressure and this may change as the cohort transition to adulthood.
Limitations
When modelling neighbourhood health effects it is common for researchers to use a multilevel approach to control for the clustering of SES characteristics within neighbourhoods (Pickett & Pearl, Reference Pickett and Pearl2001). However, the definition of neighbourhood used here (within 20 minutes of home) meant that no two households exactly shared a neighbourhood, although they will share some aspect of their neighbourhood with other participants. Not being able to correct for this partial lack of independence could lead to slightly biased standard error estimates. This study also lacks neighbourhood SES measures before age 16 years, which means that it was only possible to study neighbourhood SES in adolescence and take a cross-sectional approach to studying associations with systolic blood pressure. Cross-sectional approaches make it difficult to assess whether associations between health outcomes and neighbourhood factors are due to the context of the neighbourhood or its composition (Kawachi & Subramanian, Reference Kawachi and Subramanian2007; Do & Finch, Reference Do and Finch2008). This problem can be overcome by using a longitudinal approach that controls for pre-neighbourhood-exposure baseline characteristics (Do & Finch, Reference Do and Finch2008). As Bt20 begins to build a longitudinal approach to assessing the neighbourhood, future research will be able to incorporate such analyses. A limitation of using data from a longitudinal study is that participants often have some level of missing data. In order to maximize the sample size available for the analyses presented here some participants had missing data on some of the predictor variables. These individuals were classified as a separate missing group where relevant. The number of missing cases is small but in a number of cases this resulted in statistically significant associations being displayed between the missing group and blood pressure outcomes, indicating that the missing groups were not missing at random. If these missing cases were selectively from a particular SES group this might have changed the conclusions that were made regarding the strength of association between the SES measures and the systolic blood pressure outcomes. A final limitation of the study is the relatively small sample size, which has the power to detect medium to large effect sizes, but not small effect sizes. It is therefore only possible to conclude with a reasonable level of certainty about associations that have a medium to large effect size on systolic blood pressure. Despite this final limitation these findings confirm those observed in other adolescent studies in high and middle income settings that did analyse data from larger samples.
Conclusions
Returning to the framework that Glass & McAtee (Reference Glass and McAtee2006) propose where time (stage of life course) is on the horizontal axis and hierarchies of influence on the outcome (e.g. household and neighbourhood) are on the vertical axis, this study has shown that SES measures at any of these levels on the vertical axis account for a very small amount of the variation in systolic blood pressure in these urban South African 16 year olds. Time had little influence on the size of the effect of the SES measures on systolic blood pressure. Our findings are similar to those of other studies of the association between SES and systolic blood pressure in high income environments that have found weaker associations in adolescent populations compared with the stronger associations that have been observed in adults. This study has extended the test of the evidence in relation to systolic blood pressure/pre-hypertension beyond previous studies in two ways: 1) by focusing on adolescent household and neighbourhood SES environments in a middle income country, and 2) by using SES data from two points in the early life course (birth and age 16 years) to give a longitudinal test of the association between SES measures and systolic blood pressure. Despite the extension of the findings of previous work, there remain a number of untested components of Glass & McAtee's (Reference Glass and McAtee2006) model that future studies could address. Firstly, these associations have only been tested to age 16, where risk for increased systolic blood pressure is low in this South African cohort. Future work should extend the test of these longitudinal associations to later into the life course in middle income countries. Secondly, these associations were considered in South Africa, which is advanced in its social, economic and epidemiological transition compared with other sub-Saharan African countries and provides a unique social setting because of the apartheid legacy that this cohort were born into. Further testing of these associations is needed in other sub-Saharan African countries and other low income environments. Finally, this paper has not studied evidence for the biological factors that embody social and environmental experiences of humans and this evidence is clearly needed in low and middle income countries to add to the growing evidence base on this topic in high income countries.
Acknowledgments
The Bt20 receives financial and logistic support from the University of the Witwatersrand, and the Bone Health study was financially supported by the Wellcome Trust (UK) and Birth to Twenty continues to be supported by the Wellcome Trust (UK) (reference number 092097/Z/10/Z). The socioeconomic analysis was funded by the Medical Research Council (UK) grant id 70363 awarded to P. Griffiths. The authors would like to thank the Bt20 participants and research team, and also Gretchen Hanke for her help in cleaning the age-16 socioeconomic data.
Open access
