Article Text
PDF
Abstract
Background Inequities in the distribution of environmental exposures may add an extra burden to socially disadvantaged populations, especially when acting during vulnerable periods such as pregnancy and early life, but such inequities may be more complex and uncertain than is generally assumed. We therefore examine whether socioeconomic inequities exist in pregnancy exposures to multiple common environmental contaminants in air, water and food.
Methods A Spanish population-based birth cohort study enrolled over 2000 pregnant women between 2004 and 2008. Questionnaires assessed parental education, occupation, country of birth, diet and many other factors. Environmental pollutant assessments included nitrogen dioxide as a marker of traffic-related air pollution, trihalomethanes as a marker of tap water disinfection by-products, organochlorine biomarkers measured in maternal serum during pregnancy (polychlorinated biphenyls (PCB), dichlorodiphenyl dichloroethylene (p,p′-DDE), hexachlorobenzene and β-hexachlorocyclohexane) and mercury concentrations measured in cord blood.
Results Associations between socioeconomic status indicators and nitrogen dioxide and trihalomethanes were generally weak and inconsistent in direction. Concentrations of PCB, hexachlorobenzene and mercury were higher in higher social classes than lower social classes. p,p′-DDE and β-hexachlorocyclohexane were not related to social class. Social class explained between 1% and 5% of the variability in pollutant concentrations, much less than other variables such as region of residence, country of birth and maternal age.
Discussion This study demonstrates that the general assumption that more disadvantaged populations have higher levels of exposure to environmental pollution does not always hold and requires further elucidation in different international settings.
- Child health
- environmental exposure
- environmental health
- pollution
- pregnancy
- social inequalities
Statistics from Altmetric.com
It is well known that there are large socioeconomic disparities in pregnancy and child health outcomes1 2 and that these can influence adult health status.3 Exposure to potentially hazardous environmental pollutants may add another burden to socially disadvantaged groups already experiencing higher exposure to other risk factors such as smoking, poor diet and poor housing. Such a combined burden may result in the exacerbation of adverse health effects, especially when acting during vulnerable periods such as pregnancy and early life,4 and would therefore have important social and public health implications. In the USA, a large body of literature has developed on environmental injustice, generally defined as inequity in the distribution of environmental exposures.5 6 A general conclusion is that disadvantaged groups of society are more likely to live in areas of greater air pollution and in the vicinity of polluting industries and waste sites and may be exposed to higher levels of toxins.5–8 On the other hand, it has recently been argued that the association between pollution and socioeconomic status is likely to be complex and still not well understood, and that particularly the strength, its ubiquity outside the USA, and its application to multiple pollutants remain uncertain.9 10
What is certain is that evidence for environmental injustice outside the USA is very scarce. In Europe there is growing interest in determining the importance of environmental inequities, but few studies have examined such issues, and they have mainly been limited to traffic-related air pollution.8 9 11–13 A detailed nationwide study in England of environmental inequity in many spatially varying environmental pollution sources showed that associations are generally weak and inconsistent, and depend on the pollution source, deprivation measure and level of spatial aggregation.9 Studies linking the spatial distribution of pollutants with that of deprivation have as a main limitation that the group-level associations analysed may not reflect the individual experience. For environmental pollutants that are not strongly spatially determined, such an approach is clearly not sufficient. Organochlorine compounds, for example, are a group of lipophilic, persistent chemicals that originated from the production of pesticides and other synthetic products and have bioaccumulated in our food chain; their exposure levels are mainly determined by diet and other lifestyle habits. Very little is known about social determinants of such exposures and they have generally not been included in environmental equity assessments.14 15
Further knowledge on the socioeconomic distribution of exposure to multiple common environmental pollutants during vulnerable periods of development and in parts of the world with little available data on this topic would help to complete the environmental inequity picture. In addition, such knowledge would help the assessment of the importance of socioeconomic status as a confounding variable in associations between environmental pollution and pregnancy and child health, an often-quoted reason for caution in the interpretation of environmental epidemiological studies in this area.
In Spain, the INfancia y Medio Ambiente: Environment and Childhood (INMA) birth cohort study was set up to examine the effects of environmental pollutants on pregnancy outcome and child health. The study is unique in that it has collected estimates of exposure to multiple environmental pollutants for over 2000 pregnant women. Exposures include spatially distributed exposures (to traffic-related air pollution and water disinfection by-products) and individual biomarkers of exposure to mercury and organochlorine compounds (polychlorinated biphenyls (PCB), dichlorodiphenyl dichloroethylene (p,p′-DDE), hexachlorobenzene and β-hexachlorocyclohexane). The purpose of the current study is to examine whether socioeconomic inequalities exist in pregnancy exposures to these common water, air and food contaminants.
Methods
Study population
Population-based birth cohorts were established as part of the INMA in several regions of Spain following a common protocol.16 This analysis is based on the INMA cohorts of Gipuzkoa (Basque Country), Sabadell (Catalonia) and Valencia, established between 2004 and 2008. Pregnant women were enrolled during the first trimester of pregnancy at the primary healthcare centre or hospital, depending on the region, providing they fulfilled the inclusion criteria (age ≥16 years, intention to deliver in the reference hospital, no problems of communication, singleton pregnancy, no assisted conception).16 Information on maternal and paternal education, occupation, age, country of birth, previous pregnancies and many other variables was obtained through questionnaires administered to the mother during the first and third trimesters of pregnancy. Food frequency questionnaires were administered in the first and third trimester. Informed consent was signed and the study was approved by the hospital ethics committees in the participating regions.
Socioeconomic variables
The highest achieved level of education of the mother was categorised into three groups: first 4 years of (compulsory) secondary education or less, further secondary or vocational education (high school) and university diploma. Country of birth of the mother was categorised into Spain, Latin America and other (including mainly other European countries). Maternal social class was coded from the longest held job during the pregnancy, or, if the mother did not work during pregnancy, the last job before the pregnancy. In the few cases that the mother never worked, the last job of the father was used (n=22). Occupations were coded using the four-digit Spanish Clasificación Nacional de Ocupaciones (CNO94), which is closely related to the international ISCO88 coding system.17 Five social class categories were then created following the methodology proposed by the Spanish Epidemiological Society:18 SCI: managers of companies with 10 or more employees, senior technical staff, higher level professionals; SCII: managers of companies with less than 10 employees, intermediate level professionals; SCIII: administrative and financial management supporting personnel, other self-employed professionals, supervisors of manual workers, other skilled non-manual workers; SCIV: skilled and partly skilled manual workers; SCV: unskilled manual workers.
Assessment of environmental exposures
Air pollution
Exposure to traffic-related air pollution during pregnancy was estimated using land use regression modelling for the cohorts of Sabadell19 and Gipuzkoa, and a combination of kriging and land use regression modelling for the cohort of Valencia.20 For this purpose, levels of nitrogen dioxide (considered as a marker of motor vehicle exhaust toxins) were measured at 57 sites in Sabadell, 93 sites in Valencia and 78 sites in Gipuzkoa between 2004 and 2007, using passive samplers (Radiello, Fundazione Salvatore Maugeri, Padua, Italy). Models were then applied to predict outdoor levels of nitrogen dioxide at the women's residential addresses and adjusted for temporal variations to obtain pregnancy-specific individual exposures.21 The percentages of variability in nitrogen dioxide levels explained by the exposure models were 75% for Sabadell, 81% for Valencia and 51% for Gipuzkoa. Nitrogen dioxide levels averaged for whole pregnancy were used in this paper.
Water disinfection by-products
Total trihalomethane levels were used as a marker of disinfection by-product concentrations in tap water. Levels of trihalomethane in the study areas were ascertained through sampling campaigns and collection of regulatory data from local authorities and water companies.22 Trihalomethane was determined in 421 water samples in Gipuzkoa (own sampling), 198 in Sabadell (148 own sampling, 50 from regulatory measurements) and 134 in Valencia (own sampling). In each region, predictive generalised additive models were developed to predict average household tap water trihalomethane concentrations for each month of pregnancy, using geographical and temporal variables (month, year) as predictive variables (R2 between 45% and 81%).
Organochlorine compounds
Levels of organochlorine pollutants (hexachlorobenzene, β-hexachlorocyclohexane, dichlorodiphenyl trichloroethane (p,p′-DDT), p,p′-DDE and PCB congeners 28, 118, 138, 153 and 180) were analysed in maternal serum extracted during the first trimester of pregnancy, using gas chromatography methods described elsewhere.23 The limit of detection was 0.071 ng/ml in Sabadell and Gipuzcoa and between 0.01 and 0.071 ng/ml in Valencia. For comparison purposes, values in Valencia below 0.071 ng/ml were set as non-detected. Samples with non-detectable levels were then set at a value of half the detection limit. p,p′-DDT was not further analysed as more than 79% of values were below the limit of detection. Total PCB levels were calculated as the sum of all individual congeners, PCB138, 153 and 180 were the predominant congeners. All exposures are expressed on a lipid basis in ng/g lipid using the method described elsewhere.24 Correlations between lipid adjusted and not adjusted values were high (0.99 for p,p′-DDE and 0.97 for PCB).
Mercury
Total mercury levels were analysed in cord blood, collected using venipuncture of cord vessels before the placenta was delivered. Total mercury was analysed by thermal decomposition, amalgation and atomic absorption spectrometry with a detection limit of 2.0 μg/l. Values with non-detectable levels were then set at a value of half the detection limit. Lead levels were measured in the same way but were below the limit of detection of 5.0 μg/dl for almost all cohort members and were therefore not analysed.
Statistical analysis
All exposure variables showed non-normal distributions in graphical evaluations and normality tests (qqplot, histograms, Shapiro–Wilk test) and were therefore log-transformed; the geometric mean (GM) and the 10th and 90th percentiles were used to describe their distributions. Exposures assessed through biological markers (p,p′-DDE, PCB, hexachlorobenzene, β-hexachlorocyclohexane and mercury) showed considerable overlap between the three study regions, thus analyses were pooled across the three regions. Distributions of environmental nitrogen dioxide and trihalomethane levels showed little overlap between the regions, so pooled analyses were not conducted. Multivariate linear regression models were used to examine how average pollutant concentrations vary by categories of socioeconomic variables (social class, education and country of birth) and to calculate unadjusted and adjusted geometric means and 95% CI. As such models assume a log-linear response and this assumption is hard to verify, we additionally applied logistic regression models to examine whether the rate of subjects with ‘high’ exposure biomarker concentrations varied by socioeconomic variables. High exposure was defined as concentrations above the 75th percentile of the distribution in the entire population. Trend tests were conducted by including social class and educational categories as continuous variables in all models.
Multivariate models for air and water contaminants included indicators of urban–rural area type in the two regions where this was appropriate (Valencia and Gipuzkoa); the third region included only urban areas. In the multivariate models for biomarker levels, covariates evaluated were those reported to influence concentrations of organochlorine compounds or mercury in the literature: maternal age, pre-pregnancy body mass index, total weeks of breastfeeding related to previous pregnancies and consumption of fish estimated from first trimester food frequency questionnaires; they were retained in the final adjusted models if their effect was statistically significant (p<0.05) in the multivariate models. Region was always included in the pooled models. Social class and education models were also adjusted for country of birth, and country of birth models were adjusted for social class. Social class and education were never included in the same model because of colinearity. Biomarker models were repeated restricted to Spanish-born women as biomarker concentrations in Spanish and foreign-born women showed large differences and significant interactions with social class. Likelihood ratio tests were used to test for interaction between socioeconomic variables and other covariates. Finally, analysis of variance models were applied to calculate the proportion of the total variability in pollutant concentrations explained by the covariates identified above.
Results
A total of 2081 pregnant women from the three Spanish regions was included (table 1). Women in the highest social class (SCI) were most likely to come from the Gipuzkoa region (51%), women in the lowest social class (SCV) were more likely to come from the Valencia region (63%). The cohort consisted predominantly of Spanish-born women; foreign-born women were mainly from Latin American countries and they were most prominent in SCV (19% Latin America-born and 7% other foreign-born women in SCV, compared with 3.5% and 1.5%, respectively, in SCI). Levels of pollutants varied across the regions (table 2), especially for nitrogen dioxide and trihalomethane (table 2).
Cohort characteristics by social class (N=2081)*
Distribution of environmental exposure variables by region
Associations between socioeconomic status indicators (social class, education, country of birth) and air pollution and water disinfection by-products were generally weak and inconsistent between the different regions (table 3). The strongest trend for nitrogen dioxide was found in Valencia, where nitrogen dioxide levels increased from 33.0 μg/m3 in SCI to 36.9 μg/m3 in SCV. It should be noted that in Valencia 35% of women in SCI had nitrogen dioxide levels above the EU guideline value of 40 μg/m3, compared with 54% of women in SCV (χ2 p=0.03). In Sabadell these percentages were 20% and 10%, respectively (χ2 p=0.55). In Sabadell, 88% of women in SCV had trihalomethane levels above the guideline value of 100 μg/l, compared with 56% in SCI (χ2 p=0.02). The percentage contribution of social class to variability in nitrogen dioxide and trihalomethane levels was low: 1–2% for nitrogen dioxide and 0.3–3% for trihalomethane.
Geometric mean and 95% CI of nitrogen dioxide (μg/m3) and trihalomethane (μg/l) by social class, education and country of origin, by region
All biomarkers except β-hexachlorocyclohexane showed trends with social class before adjustment for other explanatory variables: women of SCV had higher concentrations of p,p′-DDE, but lower concentrations of PCB, hexachlorobenzene and mercury (unadjusted results shown in appendix 1 online). The p,p′-DDE trend was mostly due to the high p,p′-DDE concentrations in women from Latin American origin; in Spanish-born women there was little evidence of a trend before or after adjustment for other variables (see appendix 1, table 4). Trends for other biomarkers weakened somewhat, but did not change direction when excluding non Spanish-born women (see appendix 1). After adjustment for maternal age, country of birth and previous breastfeeding, concentrations of PCB, hexachlorobenzene and mercury remained higher in women of higher social class than in women of lower social class; PCB concentrations, for example, decreased from 125 g/g lipid in SCI to 109 g/g lipid in SCV (table 4). Similar associations were observed with maternal education. OR for high exposure reflected very similar social class patterns; women in SCV had a 0.37 lower odds of high PCB exposure compared with SCI (table 4). Country of birth was a strong predictor of organochlorine and mercury concentrations; adjusted p,p′-DDE levels were nearly three times higher in women born in Latin America compared with women born in Spain. Levels of PCB, hexachlorobenzene, β-hexachlorocyclohexane and mercury were all lower in Latin America-born women than in Spanish-born women (table 4). Women born in other countries (mainly other parts of Europe) had higher levels of p,p′-DDE and β-hexachlorocyclohexane than Spanish women, but lower levels of hexachlorobenzene and mercury. We found little evidence for interactions between region or other covariates and social class (p values for interaction terms >0.1).
Adjusted geometric mean and 95% CI of organochlorine (g/g lipid) and mercury (μg/l) concentrations by maternal social class, education, and country of birth, all regions combined
The percentage of the total variability in organochlorine and mercury concentrations explained by social class was relatively small compared with other variables, between 1% and 4% (figure 1). Country of birth contributed little to the variability in the β-hexachlorocyclohexane and mercury concentrations, but between 7% and 18% in the p,p′-DDE, PCB and hexachlorobenzene concentrations. Region was a relatively large contributor to variability, with between 4% for mercury and 18% for β-hexachlorocyclohexane. Maternal age contributed 7–12%. In total, the covariates shown in figure 1 explained between 15% (for mercury) and 46% (for PCB) of the variability in pollutant levels. Models including education instead of social class showed a smaller contribution of education and similar results for the other covariates (not shown).
Percentage contribution to total variability in log transformed concentrations of dichlorodiphenyl dichloroethylene (p,p′-DDE), polychlorinated biphenyls (PCB), hexachlorobenzene (HCB), β-hexachlorocyclohexane (β-HCH) and mercury (Hg), explained by covariates in multiple regression models. Variables were included in analysis of variance (ANOVA) models in the order shown in this figure. *p<0.05 for inclusion of the variable in the ANOVA model.
Discussion
This study demonstrates that the association between socioeconomic position and levels of exposure to environmental pollutants in pregnant women in Spain is weak and inconsistent, with some exposures more prevalent in higher and others in lower social and educational classes, and with changing patterns between regions for some exposures. Social position indicators explained a persistent, but small, part of the variability in levels of individual pollutant biomarkers. Most importantly, this study shows that the assumption that more disadvantaged groups of pregnant women have higher exposure levels does not always hold and will depend on the type of exposure, and on the location of both the early-life and current residence of the person.
Two of the exposures included in this study, traffic-related air pollution and water disinfection by-products, were estimated by geographical exposure prediction models. Only a few previous studies have explicitly investigated the relationship between social position and spatially determined exposures within Europe, observing higher levels of air pollution in more deprived populations in the European cities of Malmo,12 Oslo,25 Strasbourg13 and London,26 but higher levels in more affluent populations in Rome.27 Others have also discussed that in certain European cities, especially in southern Europe, affluent people live more commonly in city centres or towns, and that associations between deprivation and air pollution could therefore be reversed.28 The nationwide study in England by Briggs et al9 observed weak and inconsistent associations between air pollution and trihalomethane concentrations in drinking water and area deprivation measures, and concluded that evidence for environmental inequity was limited. Our study supports the argument that in the European setting, environmental inequalities in spatially determined exposures may not always be large and may not always be negative in direction (ie, lower exposures with increasing affluence). The social class coding system used in our study is closely related to the international ISCO88 coding system used commonly in European settings. However, the concept of social class in Europe may be very different from that in other parts of the world, in particular from that in the USA. This study highlights the fact that results from US studies cannot easily be translated to Europe.
The association between socioeconomic status and body concentrations of persistent environmental chemicals in the general population is poorly documented. There are surprisingly few studies that explicitly analyse social variations in organochlorine compounds in the general population,14 29 30 or in pregnant women,15 31 32 and sample sizes have generally been small. Findings include higher levels of hexachlorobenzene but not of other organochlorines with higher educational level in Sweden,31 higher PCB but not p,p′-DDE levels with higher social position in the USA,15 29 and higher PCB and p,p′-DDE levels in non-whites in the USA.30 32 In Spain, both positive and negative associations with social class have been reported, but always in small populations.14 33 34 Our study indicates higher levels of PCB and hexachlorobenzene, but not of the other organochlorine compounds, in higher social classes. An important finding is that a woman's place of birth and place of residence are more important in determining organochlorine levels than her social class or educational level. p,p′-DDE levels are well known to be high in Latin America due to higher past use of p,p′-DDT pesticides. Lower levels of the other organochlorine compounds in women from Latin America may be explained by differences in life-long dietary and lifestyle habits. Further elucidation of such influences on organochlorine body burdens is warranted.
Mercury levels in this study showed a strong inverse relationship with social class. The most common route of human exposure to mercury is thought to be fish consumption. Levels of mercury in Spain are relatively high, due to high levels of fish consumption, and commonly exceed the US Environmental Protection Agency (US-EPA) reference level of 5.8 μg/l in cord blood.35 In our study population, levels of total fish intake were only weakly associated with social class, and although residual confounding due to imperfect measurement of fish intake is likely to explain part of the social class variations, it seems unlikely to explain all. Adjustment for specific fish types did not change our results (not shown). These findings stress the need to establish mercury exposure sources further, focussing on dietary and lifestyle habits of higher social classes.
The strengths of this study are its size—this is the largest birth cohort in Europe with individual exposure assessments to multiple pollutants—its individual-level estimation of exposure to environmental pollutants and of social position, and its assessment of multiple common pollutants of concern for pregnancy and child health. The study sample was population based, but was somewhat selective: we have indications that participation was lower among women of lower education; this is unlikely to have biased the associations, but may have led to more uncertain estimates of exposure levels in lower social classes.
The choice of social economic status indicator is a topic that requires careful consideration.36 37 Socioeconomic status is a complex concept with many dimensions at the level of the individual, household, neighbourhood and larger community. Social class in this study was based on the occupation of the mother, as we had no a priori hypothesis on whether the occupation of the mother, father, or both best represented the social position of a woman in Spain. We repeated our social class analyses using the highest social class in the household; this did not change the results. We further observed that social class was a somewhat better predictor of exposure levels than education, stressing the need to examine multiple socioeconomic status indicators. A limitation of this study is that we were not able to study other household, neighbourhood or community-level socioeconomic indicators. Multilevel approaches have been proved useful in other studies to disentangle the influence of individual and neighbourhood factors.4 A more complete measure of socioeconomic status may have explained a larger part of the variability in the exposures.
Although we were able to estimate exposure to many common environmental pollutants of concern for the Spanish public, there are others that we did not measure. For example, lead concentrations in our population were low, below the detection level, and could not be included in the analysis. We focused on chemical pollutants only and did not include housing conditions, lack of amenities, or other factors that may lead to cumulative stress in disadvantaged populations.1 4 8 Furthermore, we did not include active or passive tobacco smoking, water consumption habits (of interest for trihalomethane exposure), or time–activity patterns (of interest for air pollution exposure), as we considered these beyond the scope of this article.
Implications for child health
All of the exposures studied are suspected of affecting child health and have been linked to reductions in fetal and/or childhood growth and development.1 38 39 Any evidence for exposure to multiple environmental exposures to be concentrated in lower social classes may therefore have important implications for child health. We do not find clear evidence for a multiple environmental exposure burden in groups of lower social position. Nevertheless, there is mounting evidence from other studies that social factors may modify the harmful effects on child health of a range of environmental pollutants, including air pollution,40 41 lead42 and environmental tobacco smoke,43 which emphasises the need to conduct analyses of pollution–health relationships in different social class strata. The relationship between environmental pollution and social inequalities has been emphasised as a priority research objective.4 44 45 This study shows that such relationships are complex and require further elucidation in different settings worldwide.
What is already known on this subject
It is generally assumed that socially disadvantaged groups of the population experience higher levels exposure to environmental pollution, based almost exclusively on data from the USA.
Environmental inequity may have important social and public health implications, especially if it acts during vulnerable periods such as pregnancy and early life.
There are few data on environmental inequity in Europe, for multiple pollutants, or for exposures during pregnancy.
What this study adds
This study shows that their is little evidence for environmental inequity in exposure to multiple environmental pollutants in pregnant women in Spain.
Associations between social class and pollutant exposures are weak and inconsistent, with some exposures more prevalent in higher and others in lower social classes, and with changing patterns between regions for some exposures.
The general assumption that more disadvantaged populations have greater exposure to environmental pollution does not always hold and requires further elucidation in different international settings.
Acknowledgments
The authors would particularly like to thank Ana Maria Garcia and Maria Carmen González Galarzo for their work on the coding occupations and social classes. They are also grateful to all fieldworkers for their assistance in contacting the families and administering the questionnaires. A full listing of the INMA project researchers can be found at http://www.proyectoinma.org.
References
Footnotes
Funding The INMA project is funded by grants from Instituto de Salud Carlos III (Red INMA G03/176 and CB06/02/0041) and Fundación Roger Torné Fundació Privada. The studies in the specific regions were funded by the Spanish Ministry of Health (FIS 03/1615, 04/1436, 04/1509, 04/1112, 04/1931, 05/1079, 05/1052, 06/0867, 06/1213, 07/0314, 08/1151, 09/02647), the Generalitat de Catalunya (CIRIT 1999SGR00241), the Diputación Foral de Gipuzkoa (DFG06/004), the Department of Health of the Basque Government (2005111093) and the Conselleria de Sanitat Generalitat Valenciana.
Competing interests None.
Ethics approval This study was conducted with the approval of the hospital ethics committees in the participating regions.
Provenance and peer review Not commissioned; externally peer reviewed.