Abstract
We use the United Kingdom's Born in Bradford study to investigate whether women in lower quality environments are less likely to breastfeed. We use measures of physical environmental quality (water disinfectant by‐products [DBPs], air pollution, passive cigarette smoke, and household condition) alongside socio‐economic indicators, to explore in detail how different exposures influence breastfeeding. Drawing on evolutionary life history theory, we predict that lower environmental quality will be associated with lower odds of initiating, and higher hazards of stopping, breastfeeding. As low physical environmental quality may increase the risk of adverse birth outcomes, which may in turn affect breastfeeding chances, we also test for mediation by gestational age, birthweight, head circumference, and abdominal circumference. Our sample is composed of mothers who gave birth at the Bradford Royal Infirmary in West Yorkshire between March 2007 and December 2010 for whom breastfeeding initiation data was available. Analyses were stratified by the two largest ethnic groups: White British (n = 3,951) and Pakistani‐origin (n = 4,411) mothers. After controlling for socio‐economic position, Pakistani‐origin mothers had lower chances of initiating and higher chances of stopping breastfeeding with increased water DBP exposure (e.g., OR for 0.03–0.61 vs. <0.02 μg/day dibromochloromethane exposure 0.70 [0.58–0.83], HR 1.16 [0.99–1.36]), greater air pollution exposure predicted lower chances of initiation for both ethnic groups (e.g., OR for 10 μg/m3 increase in nitrogen dioxide 0.81 [0.66–0.99] for White British mothers and 0.79 [0.67–0.94] for Pakistani‐origin mothers) but also a reduced hazard of stopping breastfeeding for White British mothers (HR 0.65 [0.52–0.80]), and exposure to household damp/mould predicted higher chances of breastfeeding initiation amongst White British mothers (OR 1.66 [1.11–2.47]). We found no evidence that physical environmental quality effects on breastfeeding were mediated through birth outcomes amongst Pakistani‐origin mothers and only weak evidence (p < 0.10) amongst White British mothers (exposure to passive cigarette smoke was associated with having lower birthweight infants who were in turn less likely to be breastfed whereas greater air pollution exposure was associated with longer gestations and in turn reduced hazards of stopping breastfeeding). Overall, our findings suggest that there is differential susceptibility to environmental exposures according to ethnicity. Although the water DBP results for Pakistani‐origin mothers and air pollution‐initiation results for both ethnic groups support our hypothesis that mothers exhibit reduced breastfeeding in poorer quality environments, several physical environmental quality indicators showed null or positive associations with breastfeeding outcomes. We consider physiological explanations for our findings and their implications for life history theory and public health policy.
Keywords: Born in Bradford, breastfeeding, pollution, smoke, socio‐economic position, trihalomethanes
Key messages.
The association between physical environment quality and breastfeeding varies by type of exposure, outcome, and ethnicity.
There is some evidence that water disinfectant by‐products and air pollution reduce breastfeeding, particularly for initiating breastfeeding and for Pakistani‐origin mothers, but some null and some positive associations between physical environmental quality and breastfeeding were also found.
Evolutionary life history theory serves as a useful framework for understanding human reproductive behaviour. It emphasises the importance of environmental quality in predicting behaviours such as parental investment and thereby shifts focus away from individual factors and towards modifiable aspects of the environment.
1. INTRODUCTION
There are many factors which impact a woman's infant feeding journey. Here, we contribute to the developing field of evolutionary public health by using evolutionary life history theory to inform an analysis of breastfeeding behaviour risk (Wells, Nesse, Sear, Johnstone, & Stearns, 2017). Life history theory predicts that environmental quality may pattern reproductive behaviour and decision making, because low‐quality environments correlate with lower access to resources and higher morbidity and mortality risk (Dickins, Johns, & Chipman, 2012; Kaplan & Gangestad, 2012; Nettle, 2010b; Quinlan, 2007; Voland, 1998; but see Baldini, 2015). Under certain assumptions, this relatively high environmental risk is thought to trigger behavioural and physiological responses that prioritise having children relatively early, having more of them, and investing relatively less in each, that is, favouring a quantity over quality reproductive strategy in order to ensure successful reproduction despite high mortality risk (Caudell & Quinlan, 2012; Nettle, 2010a). Empirical studies have consistently shown that women in harsher environments have earlier first births, more births, and a greater risk of preterm delivery and lower birthweight and/or smaller infants. This is the case in both cross‐population (Bulley & Pepper, 2017; Caudell & Quinlan, 2012; Low, Hazel, Parker, & Welch, 2008) and within‐population studies, including high‐income populations such as the United Kingdom, where overall mortality risk is relatively low, but there is still considerable within‐population variation in mortality and morbidity (Agyemang et al., 2009; Auger, Park, Gamache, Pampalon, & Daniel, 2012; Clemens & Dibben, 2017; Luo, Wilkins, & Kramer, 2006; Pearl, Braveman, & Abrams, 2001; Schempf, Strobino, & O'Campo, 2009; Virgo & Sear, 2016).
There is less evidence that harsh environments are associated with post‐natal parental investment, but reduced breastfeeding is potentially one mechanism that could have evolved to decrease parental investment in lower quality environments thereby enabling women to achieve higher fertility in such environments (throughout most of human history, at least, when shortened breastfeeding durations would have been associated with shorter birth intervals; Caudell & Quinlan, 2012; Chisholm, 1993; Nettle, 2010a; Pepper & Nettle, 2014; Quinlan, 2007). Breastfeeding may be partly influenced by unconscious responses to environmental cues, in an evolved, (previously) adaptive response. In contemporary high‐income contexts, breastfeeding is of course just one of several ways in which mothers can invest in their children, and we make no judgement of women's different infant feeding choices here. Many mothers who do not breastfeed opt to formula feed instead, which can also be considered maternal investment, but in an economic rather than a physiological sense, especially for socio‐economically disadvantaged mothers for whom formula incurs a higher financial cost (Raisler, 2000; UNICEF, 2002). There are also many other nonfeeding‐related investments parents can make, like those made in other aspects of infant care such as protection and education (Shenk, 2011). Our study focuses on breastfeeding, an important influence on maternal and infant health (Ip et al., 2007; Victora et al., 2016); by testing the impact of physical environmental conditions we contribute a valuable hypothesis to public health interventions.
Human environments are both physical and social, and it is important to account for both factors when understanding environmental influences on health and behaviour. In the United Kingdom, the link between the social environment, as measured by socio‐economic position (SEP), and breastfeeding is well established (McAndrew et al., 2012). Disadvantaged women with lower incomes, lower status jobs, and/or lower educational attainment are less likely to intend to breastfeed and less likely to initiate breastfeeding, and if they do, they tend to breastfeed for shorter durations. Current UK policy entitles women to 12 months of maternity leave with the first 6 weeks paid at average weekly earnings, then £145.18 per week thereafter (GOV.UK, 2019). The UK Equality Act 2010 defines treating a woman unfavourably because she is breastfeeding as discrimination (Maternity Action, 2014), but the law does not currently allow a simple straightforward right to breastfeeding breaks at work, although employers do have to consider health and safety issues (Health and Safety Executive, 2019; Maternity Action, 2018). With barriers to providing lactation breaks evident in large public sector organisations (Fraser, 2016), it is likely that the smaller and less formal organisations (that more socio‐economically disadvantaged mothers often work for) will be even less supportive in this regard (Heinig et al., 2006).
Socio‐economic position is not the only way that large societies are stratified. The United Kingdom is home to great diversity, with people of various ethnic backgrounds and immigration histories calling this country home (Office for National Statistics, 2012). Ethnicity and socio‐economic status can be intertwined, with some ethnic minorities also being socio‐economically disadvantaged. For example, ethnic minorities may be more likely to live in deprived areas of the United Kingdom than their White counterparts (McAndrew et al., 2012). Socio‐economic position and ethnicity do however impart different influences on breastfeeding, and socio‐economic position may have more of a beneficial effect in some ethnic groups than others. For example, Kelly et al. (2006) found that higher income levels were associated with increased odds of initiating breastfeeding amongst White and Asian mothers but that it had less of a consistent effect amongst Black mothers. Ethnicity can be considered a proxy for differing immigration histories and cultural influences and as such is an important factor to explore in relationships between environmental conditions and breastfeeding outcomes. Although socio‐economic position and ethnicity may capture women's social, cultural, and economic constraints and opportunities, there are physical aspects of environmental quality that may also influence reproduction and parenting, either because they directly influence physiology or because they act as cues to environmental quality to which women respond by changing their reproductive behaviour (not necessarily consciously).
By virtue of different environmental exposures, socio‐economic circumstances, and cultural influences, the impact of physical environmental quality on breastfeeding is likely to vary between populations. Life history theory predicts that parental investment will be reduced in harsher environmental conditions and/or when resources are scarce. However, low‐ and middle‐income countries generally have higher breastfeeding rates than high‐income countries (Victora et al., 2016) even though the environments in these contexts are in many ways “harsher.” Analyses focusing on preindustrial societies have however shown that even in these harsher contexts, there is within‐population variation in parental investment whereby breastfeeding tracks ecological stress, with mothers terminating breastfeeding sooner under conditions of warfare and famine and weaning showing a quadratic relationship with pathogen stress (Quinlan, 2007).
We have previously shown a positive association between environmental quality and breastfeeding in the United Kingdom, both when environmental quality was measured using aggregated data at Primary Care Trust level (Brown, 2014) and with individual data at the home, street, and neighbourhood level (Brown & Sear, 2017). We looked at environmental effects on breastfeeding initiation and duration of any breastfeeding in the Millennium Cohort Study (Brown & Sear, 2017), using a broad definition of environmental quality encompassing both sociocultural and physical aspects, such as how supportive and friendly people were, whether there were signs of crime and antisocial behaviour, the built environment, perceptions of cleanliness and general environmental pollution. We found that for every one‐unit increase in objectively assessed local environmental quality, mothers were 54% more likely to initiate breastfeeding (CI [1.23, 1.92]) and 14% less likely to stop breastfeeding (CI [0.77, 0.97]), even after controlling for socio‐economic position. We also found significant effect modification with more advantaged SEP having a “buffering” effect, reducing the magnitude of the consequences of adverse environmental quality. Women from high‐income households had relatively high breastfeeding initiation rates, and those with high‐status jobs were more likely to maintain breastfeeding even in harsh environmental conditions. Here, we focus more narrowly on physical aspects of the environment, such as water disinfectant by‐products (DBPs) and air pollution, to test the prediction that a poor‐quality (harsher) physical environment will negatively impact breastfeeding, alongside an assessment of the association between SEP and breastfeeding.
1.1. Physical environmental quality and breastfeeding: Proximate mechanisms and ultimate perspectives
An integrative evolutionary approach requires both proximate explanations of how a behaviour works and also an “ultimate” explanation as to why it exists (Nettle, 2011). Ultimate explanations centre on fitness consequences of a behaviour, explain why it is favoured (or not) in certain contexts, and address its evolutionary function (Scott‐Phillips, Dickins, & West, 2011). So far, we have proposed an ultimate explanation for reduced breastfeeding in lower quality environments: lower parental investment and higher fertility are adaptive in harsh environments. But what are the proximate (i.e., immediate physiological or behavioural) mechanisms that explain the relationship between harsh environments and breastfeeding? Physical aspects of the environment may directly influence maternal and child physiology, which then influences reproductive and parenting behaviours. We acknowledge that a complication is that breastfeeding is a dyadic process influenced by the infant too (Tully & Ball, 2013). There has been relatively little research explicitly linking environmental pollutant exposure with breastfeeding outcomes, but chemical compounds have been detected in breastmilk (Stefanidou, Maravelias, & Spiliopoulou, 2009), some of which are likely to have endocrine disrupting capabilities (Pedersen et al., 2013, p. 72). Hormonal disruption or toxicity can impact mammary gland development during pregnancy (Rosen‐Carole, Auinger, Howard, Brownell, & Lanphear, 2017) and also the lactation process itself. This is certainly the case for maternal smoking as it has been shown to interfere with the milk ejection reflex, reduce milk output, alter the taste and composition of breastmilk, and suppress infant appetite and increase irritability (Amir, 2001). It is possible that exposure to passive cigarette smoke and other pollutants such as those from vehicle exhaust fumes, chlorinated water, and damp and mould will have some of the same effects, albeit that the level of toxin exposure may be substantially less.
1.1.1. Adverse birth outcomes as a mediating factor
Physical aspects of the environment may also influence reproductive and parenting behaviours through indirect links. For example, an association between environmental pollutants and breastfeeding may be mediated by adverse birth outcomes. Evidence for the relationship between pollutants and adverse birth outcomes is mixed, possibly due to varying methodology, differing levels of exposure and misclassification (Poirier et al., 2015), though there does seem to be consensus that pollution can harm the developing foetus. Pollutant exposure may be linked with an increased risk of spontaneous abortion and stillbirth (Faiz et al., 2012; Waller, Swan, Delorenze, & Hopkins, 1998), but it may also increase the risk of prematurity or having a low birthweight or small for gestational age baby (Dadvand et al., 2013; Nieuwenhuijsen, Dadvand, Grellier, Martinez, & Vrijheid, 2013).
Prematurity and low birthweight can affect an infant's ability to suckle, swallow, and breathe, increasing vulnerability to feeding problems (Wambach & Riordan, 2016). Affected babies are also more likely to be separated from their mothers at birth, for example, by being moved to incubators, depriving dyads of skin‐to‐skin contact and making establishing breastfeeding more difficult. In addition to these proximate explanations, evolutionary theory predicts that parental investment is lower when offspring chances of reproducing themselves appear diminished (Heijkoop, 2010; Mann, 1995). Therefore, an ultimate perspective predicts that in order to adjust lactational investment optimally, mothers must evaluate infant health status and reproductive value (not necessarily consciously). Several studies have provided support for this hypothesis; for example, mothers of twins have been shown to bias investment towards the healthier twin (Mann, 1995), interbirth intervals are shorter following the birth of a child with a long‐term health problem (Waynforth, 2015), and mothers of low birthweight infants have been shown to wean earlier (Bereczkei, 2001).
1.2. Our predictions
The aim of this study is to test whether mothers are less likely to breastfeed in harsh environments, a prediction derived from the evolutionary framework of life history theory. We look specifically at one region in North England to answer this question and focus on small‐scale within‐population heterogeneity in physical environmental quality. In particular, we hypothesise that worse household condition (i.e., having no central heating and being exposed to damp/mould) and greater exposure to water disinfectant by‐products, air pollution, and passive cigarette smoke will negatively impact women's breastfeeding chances by reducing their odds of initiating breastfeeding and increasing their hazards of stopping breastfeeding. We further hypothesise that these aspects of the physical environment may also have indirect effects on the same breastfeeding outcomes through potentially harming foetal development resulting in mothers having smaller neonates whom they are less likely to breastfeed.
2. METHODS
2.1. Dataset
The Born in Bradford (BiB) cohort study follows the health and well‐being of over 13,500 children born at the Bradford Royal Infirmary, West Yorkshire, England, between March 2007 and December 2010. Pregnant women were primarily recruited at 26–28 weeks of gestation when attending the hospital for routine tests. There have been several waves of data collection to date. Of relevance to this study, a baseline interviewer‐administered questionnaire was completed shortly after recruitment that captured sociodemographic data; details of delivery, birthweight, and antenatal information were obtained from maternity and radiology information systems; babies had abdominal and head circumferences measured before discharge; and breastfeeding information was recorded during health worker visits and linked back to the main dataset. Further follow‐up occurred for two subcohorts—BiB1000 and ALLIN (ALLergy and INfection)—over the first 4 and 2 years of life, respectively, from which we obtained information on breastfeeding duration. An additional subsample took part in the MEchanism of the Development of ALLergy (MeDALL) study, for whom we have additional information on household condition and breastfeeding at age 4 years. We use air pollution measures collected as part of the multisite European Study of Cohorts for Air Pollution Effects (ESCAPE) project (Pedersen et al., 2013). Routine water quality monitoring data were provided by Yorkshire Water for the eight water supply zones covering the study area from January 2006 to March 2011 and exposure levels were derived by Mireille Toledano and Imperial College of Science Technology and Medicine (Smith et al., 2016). More details on the cohort, subcohorts, and data collection is available elsewhere (Smith et al., 2016; Wright et al., 2013).
We only included mothers with live births (excluding 72 mothers), and where mothers had twins or triplets, we randomly chose one child for inclusion (excluding 182 babies). For mothers with repeated pregnancies during the data collection period (2007–2010), we randomly selected one pregnancy (excluding a further 1,286 babies). These restrictions to one‐mother–one‐child data points were to ensure each mother just contributed one case to the dataset to avoid issues of clustering at the mother level. Our sample includes mothers of varying parity, not just first time mothers as some women will have given birth prior to inclusion in the study. This gave us an initial maximum usable sample size of 12,318 mother–infant dyads.
2.2. Variables
2.2.1. Ethnicity
Given that breastfeeding practices and environmental exposure may differ by ethnicity (a proxy for differing immigration histories and cultural influences), we present stratified results, focusing on the two main ethnic groups in Bradford—White British (n = 4,031) and Pakistani‐origin (n = 4,448) mothers. We also present model results for the total sample but do not attempt to interpret results for the rest of the sample, because it comprised a heterogeneous “other” ethnicity category (n = 1,541) and women who did not provide their ethnicity (n = 2,298).
2.2.2. Breastfeeding outcomes
We used two outcomes: (a) breastfeeding initiation and (b) duration of any (rather than exclusive) breastfeeding. We combined breastfeeding initiation data from health visitor records and subcohort follow‐up. This gave us initiation data for 98% of the women in our sample (n = 12,087, missing = 231).
A total of 3,737 women took part in at least one of the subcohort surveys in which duration questions were asked, of which 80% had initiated breastfeeding (n = 2,979). We were able to derive duration for 95% of these women (n = 2,827). Duration was replaced with the baby's age for the 407 mothers who were still breastfeeding at the time of their last survey (167 of whom were White British and 159 of whom were of Pakistani origin). Where mothers stopped breastfeeding between surveys, duration was coded as the age of the child in the last survey where breastfeeding was recorded as still happening (likely underestimating durations for some mothers). Mothers who initiated breastfeeding but who recorded duration as 0 days were recoded as half a day (0.02 months) to acknowledge that some transfer of breastmilk may have occurred and to differentiate these mothers from those who did not attempt breastfeeding at all.
2.2.3. Physical environmental quality indicators
All measures were coded so that higher values represented greater exposure and poorer environmental quality. Where possible we have used data on exposure during pregnancy but have had to use later exposure as proxies for some indicators.
Water DBPs
We used five water DBP indicators: total trihalomethanes, brominated trihalomethanes (subdivided into bromodichloromethane and dibromochloromethane), and chloroform. Modelled trihalomethane concentrations encompassing residential (and workplace, if relevant) address were assigned, and time‐weighted average concentrations were calculated for each mother in the study. The time‐weighting was based on the proportion of the whole pregnancy falling into each month. These time‐weighted average concentrations were then adjusted for individual water use including consumption, showering, bathing, and swimming (Smith et al., 2016) to create a personalised measure of whole pregnancy average integrated uptake (μg/day). All five indicators had positively skewed distributions and so we created tertiles of exposure based on the full sample of women.
Air pollution
As part of the ESCAPE project, 20 European study areas collected measurements of particulate matter (e.g., PM2.5 and PM10) and nitrogen oxides (NO2 and NOx). We used nitrogen oxide measures as our indicators of air pollution, as Bradford was one of the 16 ESCAPE sites that did not collect particulate matter measures (Beelen et al., 2013). In addition, the evidence is less consistent for links between nitrogen oxides and infant health outcomes (Shah & Balkhair, 2011), and as such, our paper makes an important contribution to the evidence base. Furthermore, Bradford is one of the United Kingdom's nitrogen oxide pollution hotspots (Google My Maps, 2019): Nitrogen oxide levels were relatively high in Bradford between 2007 and 2010, surpassing the annual average air quality objective level of 40 μg/m3 (Maybury, 2016), and levels have remained high in recent years (Department for Environment Food and Rural Affairs, 2019). As part of the ESCAPE project, exposure estimates were personalised with land use regression models to take into account each mother's proximity to traffic and buildings and their load and density at different time points during pregnancy (Beelen et al., 2013). We selected whole pregnancy average exposure levels of nitrogen oxides (20 μg/m3) and nitrogen dioxide (10 μg/m3) for use in the present study. The nitrogen oxides indicator encompasses nitrogen dioxide and nitric oxide. We used continuous indicators in the main models (but created tertiles when testing for interactions with ethnicity) as both measures were normally distributed.
Passive cigarette smoke
Mothers were asked in the baseline questionnaire if they were exposed to cigarette smoke at work or at home, and we collapsed Yes and Less than an hour into Yes to make this a binary variable.
Household condition
We used two binary variables for household condition based on maternal reporting of damp and/or mould and lack of central heating, derived from the ALLIN and MeDALL subcohorts at 12 and/or 24 months and/or 4 years.
2.2.4. Socio‐economic position
A wide range of SEP indicators have been shown to be associated with both adverse birth outcomes (Erickson & Arbour, 2014) and breastfeeding (McAndrew et al., 2012). As a proxy for individual resources, and to some extent, social environmental quality, we wanted to capture the multifactorial nature of socio‐economic position so used five indicator variables (all taken from the baseline questionnaire) to construct a latent variable: mother's education, her partner's occupation, financial difficulties, means‐tested benefits, and food insecurity. This allocated everyone a disadvantage score that we then standardised to aid interpretation of model results. We also included the Index of Multiple Deprivation (IMD) 2010 score (McLennan, Barnes, Noble, Davies, & Garratt, 2011) as a measure of neighbourhood deprivation in descriptive analyses. Higher values represent more disadvantage for both SEP measures.
2.2.5. Covariates
We adjusted for key maternal and infant characteristics known to influence breastfeeding and/or birth outcomes: maternal age, immigration status, smoking during pregnancy, BMI, parity, infant sex, singleton/multiple birth, and cohabitation status. This reduced our maximum sample size down to 8,993 mothers (3,615 White British and 3,982 Pakistani origin).
2.2.6. Birth outcomes
Birthweight in kilograms, head and abdominal circumferences in centimetres, and gestational age in weeks were used as continuous measures in our mediation models.
2.3. Statistical methods
We first explored the data by using t tests and chi‐squared tests to compare White British and Pakistani‐origin mothers in terms of their sociodemographic characteristics, birth outcomes, environmental exposures, and breastfeeding outcomes. We also compared those with missing initiation data to the rest of the sample in the same way. We assessed associations between physical environmental quality indicators using polychoric, polyserial, and Pearson's correlations as appropriate. Unadjusted associations between our two measures of SEP (socio‐economic disadvantage and neighbourhood deprivation) and physical environmental quality were assessed using linear and logistic regression models as appropriate. To test our hypothesis about the association between physical environmental quality and breastfeeding, we ran separate statistical models for each of the 10 physical environmental quality indicators and breastfeeding outcomes, that is, only including one indicator at a time. Our first set of models adjusted for the key maternal and infant covariates listed above, and our second set of models additionally adjusted for the standardised socio‐economic disadvantage score. This allowed us to see whether any association persisted above and beyond the effect of individual socio‐economic position (we also ran models to test the association between socio‐economic position and breastfeeding outcomes). Logistic regression models were used to assess relationships with breastfeeding initiation, whereas we used event history analysis to take account of the right‐censored breastfeeding duration data (using the Weibull distribution to reflect the diminishing probability of breastfeeding over time). Breastfeeding duration results are therefore presented as hazard ratios reflecting the risk of stopping breastfeeding for different environmental exposures. As well as running analyses separately by ethnicity, we also ran the two sets of models on White British and Pakistani‐origin mothers combined, adding in an environmental quality × ethnicity interaction to test for ethnic differences in the effect of physical environmental quality on breastfeeding outcomes. We plotted predicted probabilities of initiating and maintaining breastfeeding based on these interaction models to visually compare associations amongst the two groups. The probability of maintaining breastfeeding is presented as a survival curve where the “failure” variable is stopping breastfeeding and “surviving” is maintaining breastfeeding at a given time point.
We tested for mediation using structural equation modelling adding pathways through birth outcomes to the fully adjusted models and examining indirect effects. We ran models for birthweight, head and abdominal circumference (all simultaneously adjusted for gestational age), and for gestational age separately (with all environmental quality indicators treated as continuous to allow for estimation of indirect effects).
We conducted complete case analyses, and so sample sizes varied depending on the outcome and indicator included in the model.
2.4. Ethics
BiB and its substudies have been approved by the Bradford Research Ethics Committee (Wright et al., 2013). The current study received ethics approval from the London School of Hygiene & Tropical Medicine's Research Ethics Committee (9398–01).
3. RESULTS
3.1. Sample characteristics
White British and Pakistani‐origin mothers significantly differed in most characteristics (Table 1). Pakistani‐origin mothers had higher levels of initiation (57% vs. 42%), but duration was similar at 8–9 months for both groups. Pakistani‐origin mothers had lower exposure to water DBPs and were less likely to be exposed to passive smoke but had higher exposure to air pollution. There were no ethnic differences in either indicator of household condition. Pakistani‐origin mothers had higher SEP scores indicating greater socio‐economic disadvantage. On average, they had less education, had greater neighbourhood deprivation, were more likely to be on means‐tested benefits, and experience financial difficulties but were less likely to experience food insecurity. Pakistani‐origin mothers' partners were less likely to be nonmanual workers and more likely to be manual workers or self‐employed compared with White British mothers' partners. The immigration statuses of the two groups significantly varied; the majority of White British mothers were born in the United Kingdom, whereas almost all of the Pakistani‐origin mothers were first or second generation immigrants. Pakistani‐origin mothers had lower BMIs and higher parity but were just as likely to have a female infant or multiple birth. Fewer Pakistani‐origin mothers were living without a partner. In terms of birth outcomes, Pakistani‐origin infants tended to be born earlier and were lighter, with smaller head and abdominal circumferences.
Table 1.
Characteristics of study population
| Total (n = 12,318) | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Mothers with initiation data (n = 12,087) | Mothers missing initiation data (n = 231) | ||||||||||
| Main ethnic groups | |||||||||||
| All ethnic groups (n = 12,087) | White British mothers (n = 3,951) | Pakistani‐origin mothers (n = 4,411) | P valueb | All ethnicities | P valuec | ||||||
| Data source (time point) | n | n(%) or mean ± SD | n | n(%) or mean ± SD | n | n(%) or mean ± SD | n | n(%) or mean ± SD | |||
| Breastfeeding | |||||||||||
| Initiation | Child health records, MeDALL (4 years), ALLIN (24 and 12 m), BiB1000 (36, 24, 12, and 6 m) | 12,087 | 5,982 (49.49%) | 3,951 | 1,645 (41.64%) | 4,411 | 2,507 (56.84%) | <0.001 | 1.365 | ||
| Duration (months)a | 2,827 | 8.84 ± 8.96 | 902 | 8.21 ± 9.09 | 1,437 | 8.63 ± 8.86 | 0.271 | ||||
| Physical environmental quality | |||||||||||
| Water disinfectant by‐products | Yorkshire Water routine monitoring data (pregnancy) | ||||||||||
| Total trihalomethanes | 9,714 | 1.85 ± 1.63 | 3,863 | 2.27 ± 1.97 | 4,341 | 1.50 ± 1.21 | <0.001 | 135 | 2.05 ± 1.92 | 0.145 | |
| <1.05 μg/day | 3,244 (33.40%) | 917 (23.74%) | 1,850 (42.62%) | <0.001 | 39 (28.89%) | 0.310 | |||||
| 1.05−1.82 μg/day | 3,240 (33.35%) | 1,228 (31.79%) | 1,470 (33.86%) | 43 (31.85%) | |||||||
| 1.82−23.96 μg/day | 3,230 (33.25%) | 1,718 (44.47%) | 1,021 (23.52%) | 53 (39.26%) | |||||||
| Brominated trihalomethanes | 9,714 | 0.25 ± 0.21 | 3,863 | 0.30 ± 0.24 | 4,341 | 0.20 ± 0.17 | <0.001 | 135 | 0.29 ± 0.27 | 0.016 | |
| <0.14 μg/day | 3,244 (33.40%) | 841 (21.77%) | 1,938 (44.64%) | <0.001 | 39 (28.89%) | 0.036 | |||||
| 0.14−0.26 μg/day | 3,246 (33.42%) | 1,278 (33.08%) | 1,415 (32.60%) | 37 (27.41%) | |||||||
| 0.26−3.34 μg/day | 3,224 (33.19%) | 1,744 (45.15%) | 988 (22.76%) | 59 (43.70%) | |||||||
| Bromodichloromethane | 9,714 | 0.20 ± 0.16 | 3,863 | 0.24 ± 0.18 | 4,341 | 0.16 ± 0.13 | <0.001 | 135 | 0.23 ± 0.21 | 0.014 | |
| <0.12 μg/day | 3,244 (33.40%) | 852 (22.06%) | 1,929 (44.44%) | <0.001 | 39 (28.89%) | 0.057 | |||||
| 0.12−0.21 μg/day | 3,245 (33.41%) | 1,285 (33.26%) | 1,407 (32.41%) | 38 (28.15%) | |||||||
| 0.21−2.61 μg/day | 3,225 (33.20%) | 1,726 (44.68%) | 1,005 (23.15%) | 58 (42.96%) | |||||||
| Dibromochloromethane | 9,714 | 0.03 ± 0.03 | 3,863 | 0.04 ± 0.04 | 4,341 | 0.02 ± 0.02 | <0.001 | 135 | 0.04 ± 0.04 | 0.075 | |
| <0.02 μg/day | 3,245 (33.41%) | 900 (23.30%) | 1,862 (42.89%) | <0.001 | 38 (28.15%) | 0.057 | |||||
| 0.02−0.03 μg/day | 3,244 (33.40%) | 1,215 (31.45%) | 1,500 (34.55%) | 39 (28.89%) | |||||||
| 0.03−0.61 μg/day | 3,225 (33.20%) | 1,748 (45.25%) | 979 (22.55%) | 58 (42.96%) | |||||||
| Chloroform | 9,714 | 1.60 ± 1.44 | 3,863 | 1.96 ± 1.75 | 4,341 | 1.29 ± 1.05 | <0.001 | 135 | 1.77 ± 1.67 | 0.191 | |
| <0.91 μg/day | 3,243 (33.38%) | 933 (24.15%) | 1,831 (42.18%) | <0.001 | 40 (29.63%) | 0.332 | |||||
| 0.91−1.56 μg/day | 3,241 (33.36%) | 1,216 (31.48%) | 1,484 (34.19%) | 42 (31.11%) | |||||||
| 1.56−20.94 μg/day | 3,230 (33.25%) | 1,714 (44.37%) | 1,026 (23.64%) | 53 (39.26%) | |||||||
| Air pollution | ESCAPE (pregnancy) | ||||||||||
| Nitrogen oxides (20 μg/m3) | 9,629 | 1.80 ± 0.42 | 3,809 | 1.67 ± 0.38 | 4,313 | 1.91 ± 0.41 | <0.001 | 128 | 1.80 ± 0.43 | 0.944 | |
| <1.60 μg/m3 | 3,215 (33.39%) | 1,754 (46.05%) | 979 (22.70%) | <0.001 | 38 (29.69%) | 0.374 | |||||
| 1.60−1.95 μg/m3 | 3,212 (33.36%) | 1,308 (34.34%) | 1,432 (33.20%) | 40 (31.25%) | |||||||
| 1.95−3.81 μg/m3 | 3,202 (33.25%) | 747 (19.61%) | 1,902 (44.10%) | 50 (39.06%) | |||||||
| Nitrogen dioxide (10 μg/m3) | 9,629 | 2.14 ± 0.39 | 3,809 | 2.01 ± 0.36 | 4,313 | 2.22 ± 0.38 | <0.001 | 128 | 2.14 ± 0.40 | 0.841 | |
| <1.94 μg/m3 | 3,213 (33.37%) | 1,748 (45.89%) | 1,017 (23.58%) | <0.001 | 40 (31.25%) | 0.880 | |||||
| 1.94−2.29 μg/m3 | 3,208 (33.32%) | 1,299 (34.10%) | 1,448 (33.57%) | 44 (34.38%) | |||||||
| 2.29−3.81 μg/m3 | 3,208 (33.32%) | 762 (20.01%) | 1,848 (42.85%) | 44 (34.38%) | |||||||
| Passive cigarette smoke | Baseline (26–28 weeks of gestation) | 9,839 | 3,169 (32.21%) | 3,936 | 1,697 (43.11%) | 4,377 | 1,064 (24.31%) | <0.001 | 140 | 61 (43.57%) | 0.004 |
| Household condition | |||||||||||
| No central heating | ALLIN (24 and 12 m) | 2,198 | 123 (5.60%) | 834 | 41 (4.92%) | 1,046 | 58 (5.54%) | 0.544 | |||
| Damp and/or mould | MeDALL (4 years), ALLIN (24 and 12 m) | 2,932 | 598 (20.40%) | 1,013 | 211 (20.83%) | 1,519 | 304 (20.01%) | 0.617 | |||
| Socio‐economic position | |||||||||||
| Socio‐economic disadvantaged | BiB1000 (12 m), Baseline (26–28 weeks of gestation) | 12,087 | −0.003 ± 0.999 | 3,951 | −0.057 ± 1.146 | 4,411 | 0.146 ± 1.036 | <0.001 | 231 | 0.160 ± 1.027 | 0.014 |
| Food insecure | BiB1000 (12 m) | 1,186 | 249 (20.99%) | 443 | 128 (28.89%) | 564 | 73 (12.94%) | <0.001 | |||
| Financial difficulties | Baseline (26–28 weeks of gestation) | 10,022 | 779 (7.77%) | 3,936 | 271 (6.88%) | 4,383 | 345 (7.88%) | 0.027 | 174 | 18 (10.34%) | 0.376 |
| Means tested benefits | Baseline (26–28 weeks of gestation) | 9,863 | 3,884 (39.38%) | 3,936 | 1,431 (36.36%) | 4,398 | 2,005 (45.59%) | <0.001 | 139 | 64 (46.04%) | 0.110 |
| Partner's employment status | Baseline (26–28 weeks of gestation) | 9,369 | 3,960 | 4,214 | <0.001 | 129 | 0.037 | ||||
| Employed—Nonmanual | 3,831 (40.89%) | 1,892 (51.27%) | 1,307 (31.02%) | 44 (34.11%) | |||||||
| Employed—Manual | 3,220 (34.37%) | 1,028 (27.86%) | 1,721 (40.84%) | 43 (33.33%) | |||||||
| Self‐employed | 1,394 (14.88%) | 376 (10.19%) | 835 (19.81%) | 19 (14.73%) | |||||||
| Student | 166 (1.77%) | 53 (1.44%) | 53 (1.26%) | 3 (2.33%) | |||||||
| Unemployed | 758 (8.09%) | 341 (9.24%) | 298 (7.07%) | 20 (15.50%) | |||||||
| Education: <5 General Certificate of Secondary Education (GCSE) equivalent | Baseline (26–28 weeks of gestation) | 9,266 | 2,127 (22.95%) | 3,556 | 766 (21.54%) | 4,198 | 1,133 (26.99%) | <0.001 | 160 | 62 (38.75%) | <0.001 |
| Neighbourhood deprivation (IMD)e | Baseline (26–28 weeks of gestation) | 9,895 | 42.09 ± 17.81 | 3,949 | 36.25 ± 19.08 | 4,410 | 46.56 ± 14.79 | <0.001 | 143 | 41.80 ± 18.31 | 0.843 |
| Ethnicity | 9,879 | 3,951 | 4,410 | 141 | <0.001 | ||||||
| White British | 3,951 (39.99%) | 3,951 (100.00%) | 0 (0.00%) | 80 (56.74%) | |||||||
| Pakistani origin | 4,411 (44.65%) | 0 (0.00%) | 4,411 (100.00%) | 37 (26.24%) | |||||||
| Other | 1,517 (15.36%) | 0 (0.00%) | 0 (0.00%) | 24 (17.02%) | |||||||
| Covariates | |||||||||||
| Immigration status | Baseline (26–28 weeks of gestation) | 12,050 | 3,924 | 4,409 | 231 | <0.001 | |||||
| Born in the United Kingdom and both parents born in the United Kingdom | 3,893 (32.31%) | 3,746 (95.46%) | 25 (0.57%) | <0.001 | 82 (35.50%) | ||||||
| Second generation (at least one parent born outside United Kingdom) | 2,378 (19.73%) | 112 (2.85%) | 1,850 (41.96%) | 24 (10.39%) | |||||||
| First generation (arrived to United Kingdom as a child) | 842 (6.99%) | 46 (1.17%) | 619 (14.04%) | 7 (3.03%) | |||||||
| First generation (arrived to United Kingdom as an adult) | 4,937 (40.97%) | 20 (0.51%) | 1,915 (43.43%) | 118 (51.08%) | |||||||
| Age (years) | Baseline (26–28 weeks of gestation) | 9,898 | 27.31 ± 5.63 | 3,951 | 26.67 ± 6.08 | 4,411 | 27.70 ± 5.21 | <0.001 | 143 | 26.48 ± 6.48 | 0.078 |
| BMI | Baseline (26–28 weeks of gestation) | 9,399 | 25.97 ± 5.68 | 3,744 | 26.67 ± 5.97 | 4,188 | 25.56 ± 5.42 | <0.001 | 131 | 26.27 ± 5.55 | 0.548 |
| Smoked during pregnancy | Baseline (26–28 weeks of gestation) | 9,811 | 1,644 (16.64%) | 3,948 | 1,330 (33.69%) | 4,400 | 153 (3.48%) | <0.001 | 141 | 44 (31.21%) | <0.001 |
| Parity | eClipse (birth) | 11,381 | 1.08 ± 1.29 | 3,807 | 0.77 ± 1.03 | 4,199 | 1.37 ± 1.42 | <0.001 | 181 | 1.13 ± 1.55 | 0.656 |
| 0 | 4,979 (43.75%) | 1,996 (52.43%) | 1,521 (36.22%) | <0.001 | 87 (48.07%) | 0.002 | |||||
| 1 | 3,000 (26.36%) | 1,092 (28.68%) | 996 (23.72%) | 45 (24.86%) | |||||||
| 2 | 1,850 (16.26%) | 465 (12.21%) | 826 (19.67%) | 23 (12.71%) | |||||||
| 3 | 930 (8.17%) | 158 (4.15%) | 518 (12.34%) | 7 (3.87%) | |||||||
| 4 | 397 (3.49%) | 64 (1.68%) | 211 (5.03%) | 9 (4.97%) | |||||||
| 5+ | 225 (1.98%) | 32 (0.84%) | 127 (3.02%) | 10 (5.52%) | |||||||
| Female infant | eClipse (birth) | 11,830 | 5,731 (48.44%) | 3,942 | 1,904 (48.30%) | 4,396 | 2,142 (48.73%) | 0.698 | 191 | 90 (47.12%) | 0.717 |
| Twins or triplets | eClipse (birth) | 11,831 | 149 (1.26%) | 3,942 | 53 (1.34%) | 4,396 | 66 (1.50%) | 0.558 | 191 | 2 (1.05%) | 0.950 |
| Not living with partner | MeDALL (4 years), ALLIN (12 m), BiB1000 (12 and 6 m), Baseline (26–28 weeks of gestation) | 9,891 | 1,664 (16.82%) | 3,946 | 1,128 (28.59%) | 4,404 | 305 (6.93%) | <0.001 | 141 | 40 (28.37%) | <0.001 |
| Birth outcomes | eClipse (birth) | ||||||||||
| Gestational age (weeks) | 11,831 | 39.13 ± 1.79 | 3,942 | 39.26 ± 1.84 | 4,396 | 39.07 ± 1.73 | <0.001 | 191 | 37.76 ± 3.81 | <0.001 | |
| Birthweight (kg) | 11,830 | 3.22 ± 0.55 | 3,941 | 3.35 ± 0.56 | 4,396 | 3.13 ± 0.53 | <0.001 | 191 | 2.95 ± 0.86 | <0.001 | |
| Head circumference (cm) | 10,962 | 34.22 ± 1.62 | 3,632 | 34.52 ± 1.60 | 4,061 | 34.01 ± 1.58 | <0.001 | 138 | 34.01 ± 2.33 | 0.128 | |
| Abdominal circumference (cm) | 10,409 | 31.21 ± 2.66 | 3,447 | 31.95 ± 2.63 | 3,866 | 30.65 ± 2.59 | <0.001 | 128 | 31.53 ± 2.51 | 0.173 | |
Abbreviations: ALLIN, ALLergy and INfection subcohort; BiB1000, BiB1000 subcohort; BMI, body mass index; eClipse, maternity information system; ESCAPE, European Study of Cohorts for Air Pollution Effects; IMD, Index of Multiple Deprivation; MeDALL, Mechanisms of the Development of ALLergy subcohort.
Breastfeeding duration questions only asked of those in sub‐cohorts; 117 women with “Don't know” and 35 provided no answers to duration questions.
P values for t tests and X2 comparing White British and Pakistani‐origin mothers.
P values for t tests and X2 comparing mothers with and without missing breastfeeding initiation data.
Standardised factor score where higher scores indicate greater socio‐economic disadvantage.
IMD = 2010 Index of Multiple Deprivation score where higher scores indicate greater deprivation.
Although mothers with missing initiation data (n = 231) significantly differed from the rest of the sample in terms of some of the environmental exposures, SEP and other sociodemographic factors (Table 1), their small numbers means that these differences are unlikely to affect the interpretation of our results.
3.2. Associations between physical environmental quality indicators
In our descriptive analyses, several environmental quality indicators were positively associated with one another in both ethnic groups, although the correlations varied in strength and significance. Focusing just on significant correlations (at p < 0.05), the strongest positive correlations were between the different water DBPs (r = 0.914 to 1) and between the different air pollution indicators (r = 0.820 and 0.826). Passive smoke exposure was more weakly positively associated with water DBPs in both ethnic groups (r = 0.077 to 0.153) and also with air pollution (r = 0.057 and 0.103) and damp/mould amongst White British mothers (r = 0.125). Household condition only correlated with other environmental quality exposures amongst Pakistani‐origin mothers; both indicators were weakly positively correlated with exposure to nitrogen dioxide (r = 0.184 and r = 0.132), and having no central heating was also weakly positively correlated with nitrogen oxide exposure (r = 0.242). However, several water DBPs were negatively, albeit weakly, associated with air pollution amongst Pakistani‐origin mothers (r = −0.034 to −0.079). These correlations suggest that exposures broadly cluster together, perhaps indicating a “harsh” physical environment, particularly for White British mothers, but the separate indicators measure slightly different aspects of the environment. Correlations between the environmental quality indicators are shown in Data S1.
3.3. SEP and physical environmental quality
Disadvantaged SEP was generally associated with poorer physical environmental quality, with for example, both greater individual socio‐economic disadvantage and neighbourhood deprivation (IMD) being positively associated with greater air pollution exposure (Table 2). There were some differences by ethnicity for the other exposures though. Most notably, although White British mothers who were more socio‐economically disadvantaged and who lived in more deprived neighbourhoods had higher levels of water DBP exposure, there was no association amongst Pakistani‐origin mothers (except for between IMD and dibromochloromethane exposure). Passive smoke exposure was positively associated with socio‐economic disadvantage in both ethnic groups but with neighbourhood deprivation only amongst White British mothers. Damp/mould exposure was more likely for socio‐economically disadvantaged White British mothers and for mothers of either ethnicity living in more deprived neighbourhoods. Only Pakistani‐origin mothers living in more deprived neighbourhoods were less likely to have access to central heating. This suggests that a lower quality physical environment is broadly associated with a lower quality socio‐economic environment, though again this relationship is stronger for White British mothers.
Table 2.
Unadjusted associations between socio‐economic position and environmental quality indicators
| All mothers (n = 12,087) | White British mothers (n = 3,951) | Pakistani‐origin mothers (n = 4,411) | ||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Socio‐economic disadvantage | Neighbourhood deprivation | Socio‐economic disadvantage | Neighbourhood deprivation | Socio‐economic disadvantage | Neighbourhood deprivation | |||||||||||||||||||
| Physical environmental quality | n | Coefficient | 95% CI | P value | n | Coefficient | 95% CI | P value | n | Coefficient | 95% CI | P value | n | Coefficient | 95% CI | P value | n | Coefficient | 95% CI | P value | n | Coefficient | 95% CI | P value |
| Water disinfectant by‐products | ||||||||||||||||||||||||
| Total trihalomethanes | 9,714 | 0.038 | 0.008, 0.067 | 0.012 | 9,712 | −0.002 | −0.004, −0.000 | 0.030 | 3,863 | 0.137 | 0.083, 0.191 | <0.001 | 3,861 | 0.008 | 0.005, 0.011 | <0.001 | 4,341 | −0.027 | −0.061, 0.008 | 0.134 | 4,341 | −0.001 | −0.004, 0.001 | 0.228 |
| Brominated trihalomethanes | 9,714 | 0.010 | 0.006, 0.013 | <0.001 | 9,712 | −0.000 | −0.000, 0.000 | 0.156 | 3,863 | 0.026 | 0.019, 0.032 | <0.001 | 3,861 | 0.001 | 0.001, 0.002 | <0.001 | 4,341 | −0.001 | −0.005, 0.004 | 0.807 | 4,341 | −0.000 | −0.000, 0.000 | 0.378 |
| Bromodichloromethane | 9,714 | 0.008 | 0.005, 0.011 | <0.001 | 9,712 | 0.000 | 0.000, 0.000 | 0.429 | 3,863 | 0.021 | 0.016, 0.026 | <0.001 | 3,861 | 0.001 | 0.001, 0.001 | <0.001 | 4,341 | −0.001 | 0.005, 0.003 | 0.691 | 4,341 | −0.000 | −0.000, 0.000 | 0.463 |
| Dibromochloromethane | 9,714 | 0.001 | 0.000, 0.001 | 0.053 | 9,712 | 0.000 | 0.000, 0.000 | <0.001 | 3,863 | 0.002 | 0.001, 0.003 | <0.001 | 3,861 | 0.000 | 0.000, 0.000 | 0.021 | 4,341 | 0.000 | −0.001, 0.000 | 0.315 | 4,341 | 0.000 | 0.000, 0.000 | 0.008 |
| Chloroform | 9,714 | 0.028 | 0.002, 0.054 | 0.035 | 9,712 | −0.002 | −0.003, 0.000 | 0.024 | 3,863 | 0.112 | 0.064, 0.160 | <0.001 | 3,861 | 0.007 | 0.004, 0.010 | <0.001 | 4,341 | −0.026 | −0.056, 0.004 | 0.092 | 4,341 | −0.001 | −0.003, 0.001 | 0.212 |
| Air pollution | ||||||||||||||||||||||||
| Nitrogen oxides | 9,629 | 0.035 | 0.028, 0.043 | <0.001 | 9,626 | 0.006 | 0.005, 0.006 | <0.001 | 3,809 | 0.012 | 0.002, 0.023 | 0.025 | 3,807 | 0.001 | 0.001, 0.002 | <0.001 | 4,312 | 0.041 | 0.029, 0.053 | <0.001 | 4,312 | 0.008 | 0.007, 0.009 | <0.001 |
| Nitrogen dioxide | 9,629 | 0.038 | 0.031‐0.045 | <0.001 | 9,626 | 0.007 | 0.006, 0.007 | <0.001 | 3,809 | 0.025 | 0.016, 0.035 | <0.001 | 3,807 | 0.003 | 0.002, 0.004 | <0.001 | 4,312 | 0.039 | 0.028, 0.050 | <0.001 | 4,312 | 0.009 | 0.008, 0.010 | <0.001 |
| Passive cigarette smoke | 9,839 | 0.35 | 0.312, 0.392 | <0.001 | 9,836 | 0.010 | 0.008, 0.013 | <0.001 | 3,936 | 0.618 | 0.556, 0.679 | <0.001 | 3,934 | 0.029 | 0.026, 0.033 | <0.001 | 4,377 | 0.122 | 0.055, 0.189 | <0.001 | 4,376 | 0.000 | −0.00, 0.005 | 0.866 |
| Household condition | ||||||||||||||||||||||||
| No central heating | 2,198 | 0.13 | −0.03, 0.296 | 0.134 | 2,197 | 0.014 | 0.004, 0.025 | 0.007 | 834 | 0.255 | −0.019, 0.528 | 0.068 | 833 | 0.001 | −0.016, 0.017 | 0.919 | 1,046 | −0.032 | −0.288, 0.224 | 0.808 | 1,046 | 0.038 | 0.019, 0.058 | <0.001 |
| Damp and/or mould | 2,932 | 0.17 | 0.086, 0.253 | <0.001 | 2,931 | 0.011 | 0.006, 0.016 | <0.001 | 1,013 | 0.235 | 0.101, 0.370 | 0.001 | 1,012 | 0.011 | 0.003, 0.019 | 0.007 | 1,519 | 0.114 | −0.009, 0.237 | 0.069 | 1,519 | 0.017 | 0.009, 0.026 | <0.001 |
Note. Each row refers to a separate model, as we ran individual models for each physical environmental quality indicator. Linear regression used for all indicators except for passive cigarette smoke, no central heating, and damp and/or mould associations that were tested with logistic regression. Excludes mothers missing breastfeeding initiation data. Neighbourhood deprivation measured by the 2010 index of multiple deprivation score and socio‐economic disadvantage measured by our standardised socio‐economic position score (higher scores correspond to greater disadvantage for both measures). Positive coefficients represent predicted direction of association, whereby greater disadvantage is associated with greater environmental exposure.
Values have been bolded to highlight when p ≤ 0.05 (i.e. significant at the 5% level) and bolded and italicised when p ≤ 0.10 (i.e. significant at the 10% level).
3.4. Socio‐economic position and breastfeeding
The first lines of Tables 3 and 4 show the relationship between socio‐economic position and breastfeeding initiation and the hazard of stopping breastfeeding after controlling for maternal and infant characteristics. Although not shown for brevity, the SEP‐breastfeeding associations in the M2 physical environmental quality models were very similar to the M1 associations presented, suggesting that socio‐economic disadvantage influences breastfeeding separately from these aspects of the environment.
3.4.1. Breastfeeding initiation
More socio‐economically disadvantaged mothers had lower odds of initiating breastfeeding. The odds of initiating breastfeeding decreased by 23% and 20% for each standard deviation increase in socio‐economic disadvantage for White British and Pakistani‐origin mothers, respectively.
3.4.2. Breastfeeding duration
Mothers with greater socio‐economic disadvantage had increased hazards of stopping breastfeeding. The hazard of stopping breastfeeding increased by 11% and 13% for each standard deviation increase in socio‐economic disadvantage for White British and Pakistani‐origin mothers, respectively.
3.5. Physical environmental quality and breastfeeding
The remaining rows of Tables 3 and 4 present results of our analyses of the relationships between the physical environmental quality indicators and breastfeeding initiation and duration (hazard of stopping). M1 models are adjusted for maternal and infant characteristics, and M2 models are additionally adjusted for SEP. Note that each row in the tables refers to a separate model, as we ran individual models for each of the ten environmental quality indicators. The last column shows the results of the significance tests for the interactions between ethnic group and physical environmental quality from models controlling for maternal and infant characteristics and SEP and including White British and Pakistani‐origin mothers but excluding other ethnicities. The corresponding predicted probabilities are presented in Figure 1.
Table 3.
Associations between physical environmental quality and breastfeeding initiation
| Breastfeeding initiation | ||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| All mothers | White British mothers | Pakistani‐origin mothers | ||||||||||||||||||||
| M1: Controlling for maternal and infant characteristics | M2: Controlling for maternal and infant characteristics and socio‐economic position | M1: Controlling for maternal and infant characteristics | M2: Controlling for maternal and infant characteristics and socio‐economic position | M1: Controlling for maternal and infant characteristics | M2: Controlling for maternal and infant characteristics and socio‐economic position | |||||||||||||||||
| n | OR | 95% CI | P value | OR | 95% CI | P value | n | OR | 95% CI | P value | OR | 95% CI | P value | n | OR | 95% CI | P value | OR | 95% CI | P value | Interaction p valueb | |
| Socio‐economic disadvantagea | 8,955 | 0.776 | 0.739, 0.814 | <0.001 | — | — | — | 3,589 | 0.772 | 0.71, 0.837 | <0.001 | 0.772 | 0.71, 0.837 | <0.001 | 3,979 | — | — | — | 0.797 | 0.743, 0.854 | <0.001 | |
| Physical environmental quality | ||||||||||||||||||||||
| Water disinfectant by‐products | ||||||||||||||||||||||
| Total trihalomethanes | 8,825 | 3,517 | 3,939 | 0.103 | ||||||||||||||||||
| 1.05–1.82 μg/day versus <1.05 μg/day | 0.855 | 0.767, 0.954 | 0.005 | 0.837 | 0.750, 0.933 | 0.001 | 1.005 | 0.827, 1.221 | 0.962 | 1.005 | 0.826, 1.222 | 0.962 | 0.792 | 0.682, 0.919 | 0.002 | 0.774 | 0.666, 0.900 | 0.001 | ||||
| 1.82–23.96 μg/day versus <1.05 μg/day | 0.905 | 0.807, 1.013 | 0.083 | 0.895 | 0.798, 1.003 | 0.056 | 1.139 | 0.946, 1.370 | 0.169 | 1.150 | 0.955, 1.385 | 0.141 | 0.774 | 0.653, 0.919 | 0.003 | 0.765 | 0.644, 0.908 | 0.002 | ||||
| Brominated trihalomethanes | 8,825 | 3,517 | 3,939 | 0.360 | ||||||||||||||||||
| 0.14–0.26 μg/day versus <0.14 μg/day | 0.864 | 0.775, 0.963 | 0.008 | 0.845 | 0.758, 0.943 | 0.003 | 0.904 | 0.742, 1.101 | 0.316 | 0.901 | 0.739, 1.099 | 0.304 | 0.756 | 0.651, 0.878 | <0.001 | 0.744 | 0.640, 0.865 | <0.001 | ||||
| 0.26–3.34 μg/day versus <0.14 μg/day | 0.942 | 0.841, 1.054 | 0.297 | 0.929 | 0.829, 1.040 | 0.201 | 1.031 | 0.853, 1.246 | 0.753 | 1.045 | 0.864, 1.264 | 0.652 | 0.756 | 0.637, 0.898 | 0.001 | 0.753 | 0.633, 0.895 | 0.001 | ||||
| Bromodichloromethane | 8,825 | 3,517 | 3,939 | 0.426 | ||||||||||||||||||
| 0.12–0.21 μg/day versus <0.12 μg/day | 0.860 | 0.772, 0.959 | 0.007 | 0.841 | 0.753, 0.938 | 0.002 | 0.897 | 0.737, 1.092 | 0.279 | 0.896 | 0.736, 1.092 | 0.276 | 0.770 | 0.663, 0.895 | 0.001 | 0.756 | 0.651, 0.879 | <0.001 | ||||
| 0.21–2.61 μg/day versus <0.12 μg/day | 0.910 | 0.813, 1.020 | 0.105 | 0.900 | 0.803, 1.009 | 0.070 | 1.041 | 0.862, 1.259 | 0.675 | 1.057 | 0.874, 1.279 | 0.567 | 0.775 | 0.653, 0.920 | 0.004 | 0.768 | 0.646, 0.912 | 0.003 | ||||
| Dibromochloromethane | 8,825 | 3,517 | 3,939 | 0.018 | ||||||||||||||||||
| 0.02–0.03 μg/day versus <0.02 μg/day | 0.810 | 0.727, 0.902 | <0.001 | 0.800 | 0.717, 0.891 | <0.001 | 0.879 | 0.722, 1.069 | 0.196 | 0.881 | 0.724, 1.073 | 0.207 | 0.721 | 0.622, 0.836 | <0.001 | 0.717 | 0.617, 0.832 | <0.001 | ||||
| 0.03–0.61 μg/day versus <0.02 μg/day | 0.926 | 0.827, 1.036 | 0.180 | 0.916 | 0.818, 1.026 | 0.130 | 1.140 | 0.949, 1.370 | 0.162 | 1.150 | 0.956‐1.383 | 0.139 | 0.696 | 0.586, 0.828 | <0.001 | 0.695 | 0.584, 0.827 | <0.001 | ||||
| Chloroform | 8,825 | 3,517 | 3,939 | 0.073 | ||||||||||||||||||
| 0.91–1.56 μg/day versus <0.91 μg/day | 0.872 | 0.782, 0.972 | 0.013 | 0.853 | 0.765, 0.951 | 0.004 | 1.020 | 0.840, 1.239 | 0.840 | 1.023 | 0.841, 1.244 | 0.819 | 0.801 | 0.690, 0.930 | 0.004 | 0.782 | 0.673, 0.909 | 0.001 | ||||
| 1.56–20.94 μg/day versus <0.91 μg/day | 0.944 | 0.844, 1.057 | 0.319 | 0.933 | 0.833, 1.045 | 0.229 | 1.150 | 0.957, 1.382 | 0.137 | 1.167 | 0.969, 1.404 | 0.103 | 0.769 | 0.648,k 0.912 | 0.003 | 0.759 | 0.639‐0.901 | 0.002 | ||||
| Air pollution | ||||||||||||||||||||||
| Nitrogen oxides (20 μg/m³) | 8,751 | 0.774 | 0.694, 0.865 | <0.001 | 0.813 | 0.727, 0.909 | <0.001 | 3,477 | 0.884 | 0.729, 1.073 | 0.212 | 0.905 | 0.745, 1.100 | 0.317 | 3,907 | 0.693 | 0.593, 0.810 | <0.001 | 0.730 | 0.623, 0.854 | <0.001 | 0.441 |
| Nitrogen dioxide (10 μg/m³) | 8,751 | 0.766 | 0.682, 0.861 | <0.001 | 0.810 | 0.720, 0.911 | <0.001 | 3,477 | 0.772 | 0.631, 0.945 | 0.012 | 0.806 | 0.657, 0.988 | 0.038 | 3,907 | 0.748 | 0.633, 0.885 | 0.001 | 0.789 | 0.666, 0.935 | 0.006 | 0.846 |
| Passive cigarette smoke | 8,916 | 0.910 | 0.823, 1.007 | 0.067 | 0.966 | 0.872, 1.069 | 0.501 | 3,578 | 0.888 | 0.757, 1.041 | 0.142 | 0.961 | 0.817, 1.130 | 0.629 | 3,956 | 0.950 | 0.817, 1.104 | 0.504 | 0.992 | 0.852, 1.155 | 0.919 | 0.483 |
| Household condition | ||||||||||||||||||||||
| No central heating | 2,046 | 0.736 | 0.454, 1.193 | 0.214 | 0.787 | 0.484, 1.281 | 0.336 | 778 | 0.692 | 0.337, 1.419 | 0.315 | 0.738 | 0.359, 1.518 | 0.410 | 970 | 0.584 | 0.287, 1.186 | 0.137 | 0.624 | 0.303, 1.285 | 0.200 | 0.773 |
| Damp and/or mould | 2,725 | 1.448 | 1.116, 1.878 | 0.005 | 1.524 | 1.171, 1.983 | 0.002 | 939 | 1.570 | 1.058, 2.329 | 0.025 | 1.655 | 1.109, 2.470 | 0.014 | 1,411 | 1.344 | 0.919, 1.968 | 0.128 | 1.415 | 0.964, 2.076 | 0.076 | 0.973 |
Note. Models adjusted for cohabitation status, immigration status, body mass index, age, parity, smoking during pregnancy, the sex of the infant and whether it was a multiple birth (M1), and additionally socio‐economic position (M2). Each row in the table refers to a separate model, as we ran individual models for each physical environmental quality indicator and a separate M1 model for SEP. SEP effect sizes in each of the ten physical environmental quality M2 models were similar to SEP effect sizes in M1 models and are omitted for brevity.
Abbreviations: CI, confidence interval; OR, odds ratio.
Odds for one standard deviation increase in socio‐economic disadvantage.
P value for exposure × ethnicity interaction term in model including White British and Pakistani‐origin mothers but excluding other ethnicities.
Table 4.
Associations between physical environmental quality and breastfeeding duration
| Breastfeeding duration (hazard of stopping breastfeeding) | ||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| All mothers | White British mothers | Pakistani‐origin mothers | ||||||||||||||||||||
| M1: Controlling for maternal and infant characteristics | M2: Controlling for maternal and infant characteristics and socio‐economic position | M1: Controlling for maternal and infant characteristics | M2: Controlling for maternal and infant characteristics and socio‐economic position | M1: Controlling for maternal and infant characteristics | M2: Controlling for maternal and infant characteristics and socio‐economic position | |||||||||||||||||
| n | HR | 95% CI | P value | HR | 95% CI | P value | n | HR | 95% CI | P value | HR | 95% CI | P value | n | HR | 95% CI | P value | HR | 95% CI | P value | Interaction p valueb | |
| Socio‐economic disadvantagea | 2,635 | 1.108 | 1.060, 1.159 | <0.001 | — | — | — | 843 | 1.105 | 1.013, 1.204 | 0.024 | 1,346 | — | — | — | 1.125 | 1.058, 1.196 | <0.001 | ||||
| Physical environmental quality | ||||||||||||||||||||||
| Water disinfectant by‐products | ||||||||||||||||||||||
| Total trihalomethanes | 2,608 | 829 | 1,338 | 0.471 | ||||||||||||||||||
| 1.05–1.82 μg/day versus <1.05 μg/day | 0.986 | 0.892, 1.090 | 0.783 | 0.996 | 0.901, 1.102 | 0.942 | 0.995 | 0.814, 1.216 | 0.959 | 0.999 | 0.818, 1.222 | 0.996 | 1.046 | 0.916, 1.193 | 0.507 | 1.054 | 0.923, 1.202 | 0.437 | ||||
| 1.82–23.96 μg/day versus <1.05 μg/day | 0.991 | 0.890, 1.104 | 0.873 | 0.991 | 0.890, 1.104 | 0.871 | 0.934 | 0.768, 1.136 | 0.497 | 0.923 | 0.759, 1.123 | 0.425 | 1.123 | 0.962, 1.310 | 0.141 | 1.130 | 0.968, 1.319 | 0.122 | ||||
| Brominated trihalomethanes | 2,608 | 829 | 1,338 | 0.350 | ||||||||||||||||||
| 0.14–0.26 μg/day versus <0.14 μg/day | 0.984 | 0.889, 1.088 | 0.749 | 0.996 | 0.901, 1.102 | 0.945 | 0.976 | 0.799, 1.192 | 0.811 | 0.985 | 0.806, 1.203 | 0.882 | 1.093 | 0.956, 1.251 | 0.193 | 1.105 | 0.966, 1.264 | 0.147 | ||||
| 0.26–3.34 μg/day versus <0.14 μg/day | 1.003 | 0.900, 1.117 | 0.962 | 1.003 | 0.901, 1.118 | 0.951 | 0.910 | 0.747, 1.109 | 0.352 | 0.903 | 0.741, 1.101 | 0.313 | 1.149 | 0.984, 1.341 | 0.078 | 1.149 | 0.984, 1.342 | 0.079 | ||||
| Bromodichloromethane | 2,608 | 829 | 1,338 | 0.337 | ||||||||||||||||||
| 0.12–0.21 μg/day versus <0.12 μg/day | 0.980 | 0.886, 1.084 | 0.701 | 0.992 | 0.897, 1.098 | 0.882 | 0.964 | 0.790, 1.177 | 0.720 | 0.975 | 0.799, 1.189 | 0.800 | 1.100 | 0.961, 1.259 | 0.165 | 1.110 | 0.970, 1.270 | 0.130 | ||||
| 0.21–2.61 μg/day versus <0.12 μg/day | 1.000 | 0.898‐1.113 | 0.994 | 0.999 | 0.897, 1.113 | 0.987 | 0.905 | 0.743, 1.102 | 0.321 | 0.895 | 0.734, 1.090 | 0.271 | 1.132 | 0.971, 1.321 | 0.114 | 1.136 | 0.974, 1.325 | 0.105 | ||||
| Dibromochloromethane | 2,608 | 829 | 1,338 | 0.365 | ||||||||||||||||||
| 0.02–0.03 μg/day versus <0.02 μg/day | 1.059 | 0.958, 1.172 | 0.262 | 1.068 | 0.966, 1.182 | 0.200 | 1.061 | 0.868, 1.298 | 0.564 | 1.070 | 0.875, 1.309 | 0.509 | 1.203 | 1.053, 1.374 | 0.006 | 1.210 | 1.059, 1.382 | 0.005 | ||||
| 0.03–0.61 μg/day versus <0.02 μg/day | 1.018 | 0.915, 1.133 | 0.744 | 1.016 | 0.913, 1.131 | 0.773 | 0.920 | 0.758, 1.117 | 0.402 | 0.913 | 0.751, 1.108 | 0.356 | 1.165 | 0.996, 1.363 | 0.056 | 1.162 | 0.994, 1.360 | 0.060 | ||||
| Chloroform | 2,608 | 829 | 1,338 | 0.408 | ||||||||||||||||||
| 0.91–1.56 μg/day versus <0.91 μg/day | 0.969 | 0.876, 1.071 | 0.537 | 0.978 | 0.884, 1.081 | 0.661 | 0.959 | 0.785, 1.173 | 0.686 | 0.960 | 0.786, 1.173 | 0.691 | 1.025 | 0.899, 1.170 | 0.708 | 1.033 | 0.905, 1.178 | 0.631 | ||||
| 1.56–20.94 μg/day versus <0.91 μg/day | 0.981 | 0.881, 1.092 | 0.724 | 0.981 | 0.881, 1.092 | 0.724 | 0.919 | 0.756, 1.116 | 0.392 | 0.906 | 0.745, 1.101 | 0.321 | 1.114 | 0.955, 1.300 | 0.169 | 1.121 | 0.961, 1.308 | 0.147 | ||||
| Air pollution | ||||||||||||||||||||||
| Nitrogen oxides (20 μg/m³) | 2,580 | 0.937 | 0.841, 1.043 | 0.233 | 0.911 | 0.817, 1.015 | 0.092 | 814 | 0.786 | 0.633, 0.976 | 0.029 | 0.769 | 0.619, 0.955 | 0.018 | 1,323 | 1.051 | 0.910, 1.213 | 0.501 | 1.019 | 0.881, 1.179 | 0.799 | 0.066 |
| Nitrogen dioxide (10 μg/m³) | 2,580 | 0.833 | 0.749, 0.927 | 0.001 | 0.805 | 0.723, 0.897 | <0.001 | 814 | 0.671 | 0.540, 0.834 | <0.001 | 0.646 | 0.519, 0.804 | <0.001 | 1,323 | 0.953 | 0.825, 1.102 | 0.517 | 0.928 | 0.802, 1.074 | 0.316 | 0.034 |
| Passive cigarette smoke | 2,624 | 1.145 | 1.040, 1.260 | 0.006 | 1.120 | 1.017, 1.234 | 0.021 | 841 | 1.082 | 0.908, 1.290 | 0.378 | 1.054 | 0.883, 1.258 | 0.559 | 1,337 | 1.111 | 0.972, 1.270 | 0.124 | 1.085 | 0.948, 1.241 | 0.235 | 0.736 |
| Household condition | ||||||||||||||||||||||
| No central heating | 1,651 | 0.939 | 0.742, 1.189 | 0.601 | 0.930 | 0.734, 1.177 | 0.546 | 552 | 0.903 | 0.546, 1.493 | 0.691 | 0.898 | 0.543, 1.484 | 0.674 | 825 | 0.989 | 0.715, 1.367 | 0.946 | 0.987 | 0.714, 1.365 | 0.938 | 0.663 |
| Damp and/or mould | 2,065 | 0.953 | 0.851‐1.066 | 0.401 | 0.935 | 0.836, 1.047 | 0.245 | 662 | 0.924 | 0.747, 1.144 | 0.469 | 0.930 | 0.752, 1.151 | 0.505 | 1,070 | 1.028 | 0.882, 1.197 | 0.725 | 0.990 | 0.849, 1.154 | 0.896 | 0.578 |
Note. Models adjusted for cohabitation status, immigration status, body mass index, age, parity, smoking during pregnancy, the sex of the infant and whether it was a multiple birth (M1), and additionally socio‐economic position (M2). Each row in the table refers to a separate model, as we ran individual models for each physical environmental quality indicator and a separate M1 model for SEP. SEP effect sizes in each of the ten physical environmental quality M2 models were similar to SEP effect sizes in M1 models and are omitted for brevity.
Abbreviations: CI, confidence interval; HR, hazard ratio of stopping breastfeeding.
Odds for one standard deviation increase in socio‐economic disadvantage.
P value for exposure × ethnicity interaction term in model including White British and Pakistani‐origin mothers but excluding other ethnicities.
Figure 1.
Predicted probabilities of breastfeeding initiation (Figs. 1a, 1c, 1e, 1g, 1i) and breastfeeding duration (Figs. 1b, 1d, 1f, 1h, 1j) by physical environmental quality indicators and ethnicity. Fig. 1a and 1b ‐ dibromochloromethane (water disinfectant by‐product) exposure; Fig. 1c and 1d ‐ nitrogen dioxide (air pollution) exposure; Fig. 1e and1f‐ passive smoke exposure; Fig. 1g and 1h ‐ household central heating; Fig. 1i and 1j ‐ household damp and/or mould. Adjusted for maternal and infant characteristics and socioeconomic position. Models restricted to just White British and Pakistani‐origin mothers, with ethnicity x environmental quality interaction added.
3.5.1. Breastfeeding initiation
Results of our initiation analyses broadly suggest support for our predictions, with the exception of the damp/mould indicator, though not all of the physical environmental quality indicators were significantly associated with initiation, and there were some differences between ethnic groups.
Water DBPs
Although there were no significant relationships between DBPs and initiation amongst White British mothers, all DBP measures were significant negative predictors amongst Pakistani‐origin mothers. All five DBP measures showed a dose–response relationship whereby Pakistani‐origin mothers in the mid‐ and high‐exposure tertiles both had reduced odds compared with the low‐exposure tertiles. For example, Pakistani‐origin mothers exposed to mid‐levels of dibromochloromethane (0.02–0.03 μg/day) were 28% less likely to initiate breastfeeding compared with those exposed to low levels (<0.02 μg/day), whereas those exposed to high levels (0.03–0.61 μg/day) were 30% less likely. Adjusting for SEP had little effect on the associations between water DBPs and breastfeeding amongst Pakistani‐origin mothers; adjusted effect sizes ranged between a 22% and a 31% reduction in the odds of initiation. The significant interaction between dibromochloromethane exposure and ethnicity (p = 0.018) is shown in Figure 1a: Pakistani‐origin mothers with low exposure have a 55% probability of initiating breastfeeding, whereas those with mid and high exposure have similarly lower chances at just 47%. White British mothers show a U‐shaped relationship, with the lowest probability occurring at mid‐exposure levels (47%) and low and high levels conferring probabilities of 50% and 52%, respectively.
Air pollution
Although both air pollution measures were significantly negatively associated with breastfeeding initiation amongst Pakistani‐origin mothers, only nitrogen dioxide exposure showed a significant association amongst White British mothers. These associations persisted after adjusting for SEP, with significant effect sizes varying from a 19–27% reduction in odds of initiation. Ethnicity did not interact with air pollution to predict breastfeeding initiation (Figure 1c).
Passive cigarette smoke
Passive smoke exposure was not significantly associated with breastfeeding initiation, although relationships were in the predicted direction, with exposure to smoke at work or home conferring lower odds in both ethnic groups (Figure 1e).
Household condition
Central‐heating access showed no significant association in either group, although relationships were all in the predicted direction, whereby no access conferred lower odds of initiating breastfeeding (Figure 1g). Contrary to our predictions, damp/mould exposure showed a positive association with initiation in both groups, though this relationship was only significant in White British mothers and became stronger once SEP was controlled for, with exposed White British mothers being 66% more likely to initiate breastfeeding than those with no damp/mould exposure (Figure 1i).
3.5.2. Breastfeeding duration
In contrast to the breastfeeding initiation results, results for breastfeeding duration (hazard of stopping breastfeeding) were more mixed and did not offer strong support for our predictions.
Water DBPs
Water DBP exposure did not significantly predict hazards of stopping breastfeeding amongst White British mothers, and associations were mostly going against the predicted direction (with greater exposure predicting reduced hazards of stopping breastfeeding). Water DBP associations were however in the predicted direction for Pakistani‐origin mothers, but only dibromochloromethane was a significant predictor of the hazard of stopping breastfeeding and only at mid, not high, exposure levels. The effect became marginally stronger after controlling for SEP, with Pakistani‐origin mothers exposed to 0.02–0.03 μg/day of dibromochloromethane having a 21% higher hazard of stopping compared with those with low exposure (<0.02 μg/day). These results hint at the predicted association between higher water DBP exposure and increased hazards of stopping breastfeeding in Pakistani‐origin mothers only and suggest that dibromochloromethane may be a particular chemical of interest. Although none of the water DBPs interacted with ethnicity to predict breastfeeding duration, Figure 1b illustrates the differential impact of exposure on the two main ethnic groups clearly. Pakistani‐origin mothers with high exposure levels have lower probabilities of breastfeeding than those with low levels from about 1 month onwards, whereas the converse is true for White British mothers.
Air pollution
Contrary to our predictions, greater air pollution exposure was associated with a significant reduction in the hazard of stopping breastfeeding amongst White British mothers. Nitrogen dioxide exposure showed the same direction of association amongst Pakistani‐origin mothers, albeit non‐significantly. Effect sizes also increased after controlling for SEP, with each 20 μg/m3 increase of nitrogen oxides and each 10 μg/m3 increase of nitrogen dioxide conferring a 23% and 35% reduction in the hazard of stopping breastfeeding, respectively. The interaction between ethnicity and nitrogen dioxide exposure (p = 0.034) is clear in Figure 1d. The survival curve for nitrogen dioxide exposure shows that whilst there was no difference in the probability of maintaining breastfeeding according to exposure level for Pakistani‐origin mothers, White British mothers exposed to high levels of nitrogen dioxide (2.29–3.10 10 μg/m3) had much higher chances of maintaining breastfeeding than those exposed to low levels (<1.94 10 μg/m3), particularly after the first month or so. For example, White British mothers with high levels of nitrogen dioxide exposure have a 60% chance of breastfeeding until 6 months whereas those with low levels of exposure have just a 45% chance (Figure 1d).
Passive cigarette smoke
Passive smoke exposure did not significantly predict the hazard of stopping breastfeeding, whether or not SEP was controlled for, although associations were in the predicted direction in both ethnic groups with mothers exposed to smoke at work or at home having greater hazards of stopping breastfeeding than those unexposed. We found no significant interaction between ethnicity and passive smoke, but the corresponding survival curve suggests that smoke exposure had more of a detrimental impact on maintaining breastfeeding for Pakistani‐origin mothers than White British mothers (Figure 1f).
Household condition
Neither of the household condition indicators significantly predicted the hazard of stopping breastfeeding, before or after controlling for SEP. Relationships for central heating were in the opposite direction to that predicted with mothers without household heating having lower hazards of stopping breastfeeding compared with those with heating in both ethnic groups. Household damp/mould exposure was also nonsignificantly associated with a reduced hazard of stopping breastfeeding in the models controlling for SEP. Neither household indicator significantly interacted with ethnicity but the survival curves in Figures 1h,j suggest that worse household condition conferred higher probabilities of maintaining breastfeeding amongst the White British mothers only.
3.5.3. Mediation by birth outcomes
We found some weak evidence for mediation amongst White British mothers, with indirect effects only significant at the 10% level, but no evidence for mediation amongst Pakistani‐origin mothers.
Although there was no direct effect on breastfeeding, our mediation analyses showed that amongst White British mothers, passive smoke exposure had an indirect effect on breastfeeding initiation through birthweight: mothers exposed to passive cigarette smoke at home or at work had lower birthweight infants who were in turn less likely to be breastfed. Although this indirect effect was in the predicted direction, greater air pollution exposure (both nitrogen oxides and nitrogen dioxide) was associated with longer gestations and in turn reduced hazards of stopping breastfeeding (i.e., longer durations), which goes against our prediction that increased exposure leads to smaller neonates and reduced breastfeeding.
4. DISCUSSION
We predicted that mothers with poorer physical environmental quality, that is, greater exposure to environmental pollutants and worse household condition would be less likely to initiate breastfeeding and have higher hazards of stopping breastfeeding. We found mixed associations between physical environmental quality indicators and breastfeeding outcomes, with the direction and strength of relationship varying by indicator and ethnicity; broadly, though, the relationships with breastfeeding initiation showed stronger support for our hypothesis than those with breastfeeding duration (the hazard of stopping breastfeeding). Relationships were, perhaps surprisingly, little affected by the inclusion of socio‐economic position in models. This suggests that physical environmental quality and socio‐economic position may be separate axes of influence on breastfeeding, with some aspects of physical environmental quality impacting breastfeeding above and beyond the well‐established social and economic barriers.
4.1. Ethnic differences in breastfeeding outcomes
Pakistani‐origin mothers had higher breastfeeding initiation rates and longer average breastfeeding durations than White British mothers in our sample. It could be that varying cultural influences contribute to this difference, with for example protective Islamic beliefs (Williamson & Sacranie, 2012; Zaidi, 2014), South Asian cultural teachings (Choudhry & Wallace, 2012) and more extensive support networks (GOV.UK, 2018) amongst Pakistani‐origin mothers playing a key role. In addition, as Table 1 shows, only a small proportion of Pakistani‐origin mothers in our sample were born in the UK; our measures of physical environmental quality and socio‐economic disadvantage do not capture earlier life exposure and it may be that Pakistani‐origin mothers' breastfeeding outcomes would be better predicted by earlier exposure in Pakistan than by contemporary exposure in the UK.
Acculturation also influences breastfeeding practices; it can be thought of as the extent to which people from one culture adapt their behaviour to reflect the norms of another cultural group. This may explain the relatively similar breastfeeding durations between the two ethnic groups in our study. The detrimental influence of UK societal norms on immigrant breastfeeding chances have been reported quantitatively for immigrants generally (Hawkins, Lamb, Cole, & Law, 2008) and qualitatively for those immigrating from South Asian countries specifically (Choudhry & Wallace, 2012). For example, previous analyses using the Millennium Cohort Study have shown that although babies of South Asian descent had similar odds of being breastfed to White babies, immigrant mothers were less likely to initiate breastfeeding the longer they lived in the UK (Brown & Sear, 2017 [Supplementary material: Table 3]; Hawkins et al., 2008).
4.2. Reduced breastfeeding in harsh environments
The air pollution breastfeeding initiation results for both ethnic groups and the water DBP results for both breastfeeding outcomes for Pakistani‐origin mothers support our hypothesis that mothers are less likely to breastfeed in poorer‐quality environments. Although we did not find associations between passive smoke exposure and breastfeeding in our study (in part likely due to controlling for maternal smoking), links between smoking and breastfeeding in the literature may provide clues as to how air pollution could directly impact breastfeeding initiation. For example, air pollution may have a similar negative impact on milk ejection, output, taste and composition as well as on infant irritability and appetite (Amir, 2001). Our water DBP findings are in line with previous BiB research that found trihalomethane exposure was negatively associated with birthweight but only in Pakistani‐origin infants (Smith et al., 2016). Although we found no evidence for mediation by birthweight, together Smith's study and ours suggest that Pakistani‐origin mothers are particularly vulnerable to DBPs even though their exposure is lower. DBPs concentrate in fatty tissues, accumulating over the life course and mobilising during gestation and lactation (Colborn, Vom Saal, & Soto, 1993; Freire et al., 2011). The rate of elimination depends on the amount of fat a person has (World Health Organisation, 2005) and, as South Asian populations have more fat mass than Europeans (Deurenberg, Deurenberg‐Yap, & Guiricci, 2002; Stanfield, Wells, Fewtrell, Frost, & Leon, 2012), the Pakistani‐origin mothers in our sample may have retained DBPs in their bodies for longer, resulting in greater physiological impact. The compounds may impact breast development and lactation (Bielmeier, Best, & Narotsky, 2004; Rosen‐Carole et al., 2017) and may transfer from mother to infant (Batterman, Zhang, Wang, & Franzblau, 2002), potentially altering the taste and acceptability of breastmilk (Office of Environmental Health Hazard Assessment California Environmental Protection Agency, 2010). Alternatively, rather than being driven by physiological variation, the observed ethnic differences could be explained by social factors not controlled for, for example, diet and stress (Smith et al., 2016), or even fertility (although further data exploration shows this to be unlikely [results available on request]). Differences in childhood experiences (Belsky, 2012; Hartman, Li, Nettle, & Belsky, 2017) may also account for the ethnic differences in our results to some extent.
4.3. Breastfeeding as protection from environmental harm?
Our finding that White British mothers had reduced hazards of stopping breastfeeding (i.e., longer durations) when exposed to more air pollution is contrary to our prediction but could perhaps reflect mothers using breastfeeding to protect their infants from environmental harm. Breastfeeding provides greater antioxidative protection than formula feeding (Shoji & Koletzko, 2007) and may counteract some of the detrimental health impacts of prenatal exposure to environmental contaminants (Guxens et al., 2012), such as respiratory problems (Naz, Page, & Agho, 2016) and impaired motor and cognitive development (Lertxundi et al., 2015). Similarly, the positive damp/mould‐initiation relationship could also be explained by the protective effect of breastfeeding against the associated risks of asthma and allergies (Flohr et al., 2018; Klopp et al., 2017; Lodge et al., 2015; Silvers et al., 2012; Sonnenschein‐van Der Voort et al., 2012; Tischer et al., 2011). It is however not clear why only White British mothers showed these associations.
4.4. Mediation by birth outcomes
Although we did not find evidence that passive smoke exposure impacts breastfeeding directly, our mediation results suggest that it may restrict foetal growth, manifesting as lower birthweight with the knock‐on effect of mothers being less likely to initiate breastfeeding for these smaller infants. Compounds in cigarette smoke may cause oxidative stress to the foetal‐placental unit (Erickson & Arbour, 2014) resulting in smaller neonates.
Our finding that some of the association between greater air pollution exposure and reduced hazards of stopping breastfeeding was mediated through longer gestational lengths, although counter to our prediction of greater exposure, smaller neonates, and lower breastfeeding chances, is echoed to some extent by findings from other studies. For example, air pollution studies in Italy (Sabatino et al., 2015) and Australia (Jalaludin et al., 2007) found that greater exposure to air pollutants was associated with a reduced risk of having a preterm birth. The specific results for nitrogen oxide exposure in these studies are however more mixed: e.g., greater nitrogen dioxide exposure during the first trimester was associated with reduced preterm birth risk in the Australian study but greater exposure in the second trimester was associated with increased preterm birth risk in the Italian study. Mixed evidence for associations between nitrogen oxides and birth outcomes notwithstanding (Shah & Balkhair, 2011), it is possible that a longer gestation could serve as a mechanism by which to compensate for maternal hypoxemic‐hypoxic damage (Sabatino et al., 2015).
Taken together, these mediation results suggest that mothers with low birthweight and shorter gestation lengths have reduced breastfeeding chances, a finding corroborated by our previous analyses of the Millennium Cohort Study (Brown & Sear, 2017) that showed that lower birthweight infants had lower initiation rates and average breastfeeding durations (e.g., 67% and 2.07 months vs. 69% and 2.69 months for normal weight and 74% and 3.11 months for heavy weight). Low birthweight and preterm birth (i.e., when gestational lengths are shorter than 37 weeks) may negatively impact breastfeeding in several ways. Mothers of low birthweight infants often experience difficulties that are not common to women giving birth to healthy full‐term infants. For example, some of the underlying causes of preterm birth (e.g., hypertension, diabetes, and maternal obesity) negatively influence breastmilk production (De Freitas, Lima, Carlos, Priore, & Do Carmo Castro Franceschini, 2016). Recovering from a complicated pregnancy or delivery, feeling tired of depressed after prolonged hospitalisation, or feeling anxious due to the baby's real or apprehended condition and mourning a twin are also additional risk factors (Lefebvre, 1990; Tommy's, 2019). The baby may be more likely to be separated from the mother to be taken to the intensive care unit or to have various tests and treatments (Adamkin, 2006; De Freitas et al., 2016; Dodrill, 2011; Lefebvre, 1990). In terms of infant factors, low birthweight or preterm babies are more likely to be part of a twin or triplet set, to be sleepier and have less stamina, and exhibit signs of weakness including extreme immaturity and thermal instability and illness (including critical conditions on respirators; Adamkin, 2006; Lefebvre, 1990). These small babies are more likely to suck poorly (with immature or dysfunctional sucking skills and poor suck‐swallow‐breathe coordination) or to refuse the breast; they are less likely to be discharged exclusively breastfeeding, with mothers more likely to report feeding problems after discharge too (Dodrill, 2011; Lefebvre, 1990; Ross & Browne, 2013; UCSF Children's Hospital, 2004).
4.5. Breastfeeding initiation versus breastfeeding duration
Although breastfeeding initiation results were broadly in line with our predictions, our breastfeeding duration (hazard of stopping breastfeeding) results were more inconsistent. This could be because duration was measured on a smaller, less representative sample, with less power to detect effects. However, we also found stronger initiation than duration results in our previous study (Brown & Sear, 2017), suggesting that initiation may be genuinely more strongly influenced by environmental quality than duration. It is likely that breastfeeding duration is more influenced by other factors such as women needing to return to work (Andrew & Harvey, 2011; Heck, Braveman, Cubbin, Chávez, & Kiely, 2006; Huang & Yang, 2015; Kimbro, 2006; Rippeyoung & Noonan, 2012). It is also interesting that Pakistani‐origin mothers did not breastfeed for longer than White British mothers in this sample, even though their initiation rates were higher. It could be that associations would be more pronounced or consistent if we had used duration of exclusive breastfeeding rather than of any breastfeeding.
4.6. Implications for life history theory
Although we find some evidence for predicted associations between lower physical environmental quality and reduced breastfeeding, we also find several null associations. Moreover, the air pollution‐duration and damp/mould‐initiation associations amongst White British mothers suggest that investment may actually be increased in response to environmental risk. This could be adaptive in this low infant mortality and fertility context where replacement of infants is unlikely; it may be beneficial to invest at a higher rate and protect infants as much as possible from morbidity risk. We found some (albeit weak) evidence for mediation by birth outcomes, suggesting that mothers may be using infant viability cues to tailor investment through breastfeeding.
The different measures of physical environmental quality support interpretations of both greater and lesser lactational investment in response to environmental stressors in our study. These mixed associations might suggest that the Bradford environment is not “harsh” enough to enforce the same maternal investment decisions mothers make in environments with greater extrinsic morbidity and mortality risk. We might expect to see more pronounced and consistent reductions where environmental adversity is greater, both in terms of the measures explored in this study and in terms of other aspects of the physical environment. Quinlan's study exploring aspects such as famine and warfare provide good support for the life history theory prediction of reduced maternal investment in such harsher conditions (Quinlan, 2007). As well as the earlier weaning findings we mentioned in our introduction, his analysis of data from 186 preindustrialised societies also found that maternal care was reduced in harsher conditions (Quinlan, 2007). It is also possible that forms of parental investment other than (or instead of) breastfeeding may be reduced in poorer quality environments within high‐income contexts too, and this could prove a fruitful avenue for further research. Even though we found just limited support for our predictions, by using an evolutionary approach, we can recognise that both biology and behaviour respond to environmental cues and that “adverse” outcomes can sometimes be understood as the result of optimising reproductive strategies in a given context. We make a theoretical contribution by showing that how environmental quality is operationalised in life history theory research is important. Our mixed findings suggest that social and economic proxies of environmental quality may be more strongly linked to life history outcomes than physical measures. Even so, associations between SEP and physical environmental quality differ by ethnic group, suggesting that SEP cannot be used as a reliable proxy for environmental exposure. This nuance cautions against using socio‐economic position, environmental quality, and ethnicity interchangeably when assessing the association between environmental harshness and reproductive outcomes.
4.7. Limitations
We had to use data from time points as late as 4 years after birth for the two household condition indicators and to derive some breastfeeding information. For most women, this would have been well after they stopped breastfeeding. We have had to assume that this exposure was the same as during pregnancy, but this may not be the case for all participants. Additionally, for some indicators we only had data available for subcohorts, reducing our sample size for analysis but also the representativeness of our findings. For example, damp/mould exposure and lack of central heating access were both measured at least 12 months after birth and only for mothers in the ALLIN and MeDALL subcohorts; this might partially explain their protective and null effects, respectively, as household condition may have changed over time and the relatively smaller sample sizes may have skewed associations. A further limitation was our restricted exploration of other ethnicities due to small numbers and heterogeneity in the “other” category.
Although we were able to demonstrate some associations between physical environmental quality and breastfeeding outcomes, an understanding of the proximate mechanisms that drive these associations is needed to determine whether breastfeeding is causally associated with environmental quality. Data on potential physiological mechanisms would be particularly helpful, for example, measuring uptake and lactational transfer of pollutants. Although water DBPs have been shown to transfer to breastmilk, the amount of these chemical compounds that an infant digests will vary according to the timing of maternal exposure and the timing of feeds (Batterman et al., 2002), with different concentrations in the breastmilk likely affecting taste and acceptability to the infant to varying extents. Air pollution exposure may similarly alter the composition of breastmilk (Cinar, Ozdemir, Yucel, & Ucar, 2011). It is possible that some aspects of the physical environment are more perceivable than others, with, for example, air pollution being more detectable than the concentrations of DBPs in water. The extent to which mothers consciously detect these exposures and the extent to which they consciously adjust their breastfeeding behaviour accordingly remains to be investigated.
5. CONCLUSION
We hypothesised that poor physical environmental quality would either directly or indirectly negatively impact the breastfeeding chances of mothers in Bradford. Our predictions were only partially supported with the size and direction of associations varying according to environmental exposure, ethnicity, and breastfeeding outcome, with little evidence for an indirect effect through neonate size. From a policy perspective, in order to improve the health of the population it is important to understand how individual attributes interact with environmental exposures to produce synergistic and modifiable effects (Erickson & Arbour, 2014). The results of our study suggest that environmental hazard exposure is not always synonymous with socio‐economic disadvantage and that although the latter may be a robust predictor of lower breastfeeding chances, poor physical environmental quality has less of a consistent effect, though we did find some associations. White British and Pakistani‐origin mothers had different breastfeeding and environmental experiences even though they lived in the same geographical area, additionally highlighting the importance of ethnicity, immigration, and sociocultural influences. The impact of water DBP exposure on breastfeeding was particularly pronounced for the Pakistani‐origin mothers in our sample and we suggest that focusing on reducing the amount of chemical compounds in water (and more research into the physiological impacts of dibromochloromethane in particular) should be a public health concern. Despite the possibility of harm from environmental contaminants in breastmilk, breastfeeding is still recommended as the safest and healthiest infant feeding method. Although women should be provided with personalised infant feeding support, we suggest that it is also important to focus on tackling environmental inequities in order to facilitate successful breastfeeding.
CONFLICTS OF INTEREST
The authors declare that they have no conflict of interest.
CONTRIBUTIONS
LJB conceived the study and designed and conducted the analysis. Both authors were involved in the interpretation of the data. LJB wrote the initial draft of the manuscript and both authors contributed to revisions. Both authors approved the final version of the manuscript.
Supporting information
Data S1 Supporting information
ACKNOWLEDGMENTS
Born in Bradford is only possible because of the enthusiasm and commitment of the Children and Parents in BiB. We are grateful to all the participants, health professionals, and researchers who have made Born in Bradford happen. We thank Mireille Toledano and Imperial College of Science Technology and Medicine for the derivation of Born in Bradford disinfection by‐product exposure estimates from water supply zone boundaries and routine water quality monitoring data provided by Yorkshire Water, UK. We also thank Rob Beelen and colleagues from the European Study of Cohorts for Air Pollution Effects for collection of air pollution data and derivation of individualised exposure measures. Thank you also to three anonymous reviewers who provided helpful feedback on the manuscript.
Brown LJ, Sear R. Are mothers less likely to breastfeed in harsh environments? Physical environmental quality and breastfeeding in the Born in Bradford study. Matern Child Nutr. 2019;15: e12851 10.1111/mcn.12851
REFERENCES
- Adamkin, D. H. (2006). Feeding problems in the late preterm infant. Clinics in Perinatology, 33(4), 831–837. 10.1016/j.clp.2006.09.003 [DOI] [PubMed] [Google Scholar]
- Agyemang, C. , Vrijkotte, T. G. M. , Droomers, M. , van der Wal, M. F. , Bonsel, G. J. , & Stronks, K. (2009). The effect of neighbourhood income and deprivation on pregnancy outcomes in Amsterdam, The Netherlands. Journal of Epidemiology and Community Health, 63(9), 755–760. 10.1136/jech.2008.080408 [DOI] [PubMed] [Google Scholar]
- Amir, L. H. (2001). Maternal smoking and reduced duration of breastfeeding: A review of possible mechanisms. Early Human Development, 64, 45–67. 10.1016/S0378-3782(01)00170-0 [DOI] [PubMed] [Google Scholar]
- Andrew, N. , & Harvey, K. (2011). Infant feeding choices: Experience, self‐identity and lifestyle. Maternal & Child Nutrition, 7(1), 48–60. 10.1111/j.1740-8709.2009.00222.x [DOI] [PMC free article] [PubMed] [Google Scholar]
- Auger, N. , Park, A. L. , Gamache, P. , Pampalon, R. , & Daniel, M. (2012). Weighing the contributions of material and social area deprivation to preterm birth. Social Science & Medicine, 75(6), 1032–1037. 10.1016/j.socscimed.2012.04.033 [DOI] [PubMed] [Google Scholar]
- Baldini, R. (2015). Harsh environments and “fast” human life histories: What does the theory say? BioRxiv. bioRxiv. 10.1101/014647 [DOI] [Google Scholar]
- Batterman, S. , Zhang, L. , Wang, S. , & Franzblau, A. (2002). Partition coefficients for the trihalomethanes among blood, urine, water, milk and air. Science of the Total Environment, 284(1–3), 237–247. 10.1016/S0048-9697(01)00890-7 [DOI] [PubMed] [Google Scholar]
- Beelen, R. , Hoek, G. , Vienneau, D. , Eeftens, M. , Dimakopoulou, K. , Pedeli, X. , … de Hoogh, K. (2013). Development of NO2 and NOx land use regression models for estimating air pollution exposure in 36 study areas in Europe—The ESCAPE project. Atmospheric Environment, 72, 10–23. 10.1016/j.atmosenv.2013.02.037 [DOI] [Google Scholar]
- Belsky, J. (2012). The development of human reproductive strategies: Progress and prospects. Current Directions in Psychological Science, 21(5), 310–316. 10.1177/0963721412453588 [DOI] [Google Scholar]
- Bereczkei, T. (2001). Maternal trade‐off in treating high‐risk children. Evolution and Human Behavior, 22(3), 197–212. 10.1016/S1090-5138(01)00062-9 [DOI] [PubMed] [Google Scholar]
- Bielmeier, S. R. , Best, D. S. , & Narotsky, M. G. (2004). Serum hormone characterization and exogeneous hormone rescue of bromodichloromethane‐induced pregnancy loss in the F344 Rat. Toxicological Sciences, 77, 101–108. 10.1093/toxsci/kfh017 [DOI] [PubMed] [Google Scholar]
- Brown, L. J. (2014). Area‐level deprivation and breastfeeding in England: A quantitative analysis at primary care trust level. London School of Hygiene and Tropical Medicine. 10.13140/RG.2.2.20169.34407 [DOI] [Google Scholar]
- Brown, L. J. , & Sear, R. (2017). Local environmental quality positively predicts breastfeeding in the UK's Millennium Cohort Study. Evolution, Medicine, and Public Health, 2017(1), 120–135. 10.1093/emph/eox011 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Bulley, A. , & Pepper, G. V. (2017). Cross‐country relationships between life expectancy, intertemporal choice and age at first birth. Evolution and Human Behavior, 38(5), 652–658. 10.1016/j.evolhumbehav.2017.05.002 [DOI] [Google Scholar]
- Caudell, M. A. , & Quinlan, R. J. (2012). Resource availability, mortality, and fertility: a path analytic approach to global life‐history variation. Human Biology, 84(2), 101–125. 10.3378/027.084.0201 [DOI] [PubMed] [Google Scholar]
- Chisholm, J. (1993). Death, hope, and sex: Life‐history theory and the development of reproductive strategies. Current Anthropology, 34(1), 1–24. Retrieved from http://www.jstor.org/stable/2743728. 10.1086/204131 [DOI] [Google Scholar]
- Choudhry, K. , & Wallace, L. M. L. (2012). “Breast is not always best”: South Asian women's experiences of infant feeding in the UK within an acculturation framework. Maternal & Child Nutrition, 8(1), 72–87. 10.1111/j.1740-8709.2010.00253.x [DOI] [PMC free article] [PubMed] [Google Scholar]
- Cinar, N. , Ozdemir, S. , Yucel, O. , & Ucar, F. (2011). In which regions is breast‐feeding safer from the impact of toxic elements from the environment? Bosnian Journal of Basic Medical Sciences, 11(4), 234–239. 10.17305/bjbms.2011.2556 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Clemens, T. , & Dibben, C. (2017). Living in stressful neighbourhoods during pregnancy: An observational study of crime rates and birth outcomes. European Journal of Public Health, 27(2), 197–202. 10.1093/eurpub/ckw131 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Colborn, T. , Vom Saal, F. S. , & Soto, A. M. (1993). Developmental effects of endocrine‐disrupting chemicals in wildlife and humans. Environmental Health Perspectives, 101(5), 378–384. 10.1289/ehp.93101378 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Dadvand, P. , Parker, J. , Bell, M. L. , Bonzini, M. , Brauer, M. , Darrow, L. A. , … Woodruff, T. J. (2013). Maternal exposure to particulate air pollution and term birth weight: A multi‐country evaluation of effect and heterogeneity. Environmental Health Perspectives, 121(3), 367–373. 10.1289/ehp.1205575 [DOI] [PMC free article] [PubMed] [Google Scholar]
- De Freitas, B. A. C. , Lima, L. M. , Carlos, C. F. L. V. , Priore, S. E. , & Do Carmo Castro Franceschini, S. (2016). Duration of breastfeeding in preterm infants followed at a secondary referral service. Revista Paulista de Pediatria, 34(2), 189–196. 10.1016/j.rppede.2016.02.010 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Department for Environment Food and Rural Affairs . (2019). Site Data: Latest Data for Bradford Mayo Avenue. Retrieved May 2, 2019, from https://uk-air.defra.gov.uk/data/site-data?f_site_id=BDMA&view=last_hour
- Deurenberg, P. , Deurenberg‐Yap, M. , & Guiricci, S. (2002). Asians are different from Caucasians and from each other in their body mass index/body fat per cent relationship. Obesity Reviews, 3(3), 141–146. 10.1046/j.1467-789X.2002.00065.x [DOI] [PubMed] [Google Scholar]
- Dickins, T. E. , Johns, S. E. , & Chipman, A. (2012). Teenage pregnancy in the United Kingdom: A behavioral ecological perspective. Journal of Social, Evolutionary, and Cultural Psychology, 6(3), 344–359. Retrieved from. http://eprints.mdx.ac.uk/9442/, 10.1037/h0099247 [DOI] [Google Scholar]
- Dodrill, P. (2011). Feeding difficulties in preterm infants. Childhood Obesity and Nutrition. 10.1177/1941406411421003 [DOI] [Google Scholar]
- Erickson, A. C. , & Arbour, L. (2014). The shared pathoetiological effects of particulate air pollution and the social environment on fetal‐placental development. Journal of Environmental and Public Health, 2014, 1–20. 10.1155/2014/901017 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Faiz, A. S. , Rhoads, G. G. , Demissie, K. , Kruse, L. , Lin, Y. , & Rich, D. Q. (2012). Ambient air pollution and the risk of stillbirth. American Journal of Epidemiology, 176(4), 308–316. 10.1093/aje/kws029 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Flohr, C. , Henderson, A. J. , Kramer, M. S. , Patel, R. , Thompson, J. , Rifas‐Shiman, S. L. , … Oken, E. (2018). Effect of an intervention to promote breastfeeding on asthma, lung function, and atopic eczema at age 16 years. JAMA Pediatrics, 172(1), e174064 10.1001/jamapediatrics.2017.4064 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Fraser, M. (2016). Manager's persepctives on lactation breaks: The context of infant feeding decisions among staff in one public sector organisation. University of the West of England, Bristol. Retrieved from http://eprints.uwe.ac.uk/27471/12/Final PhD.pdf
- Freire, C. , Amaya, E. , Fernández, M. F. , González‐Galarzo, M. C. , Ramos, R. , Molina‐Molina, J. M. , … Olea, N. (2011). Relationship between occupational social class and exposure to organochlorine pesticides during pregnancy. Chemosphere, 83(6), 831–838. 10.1016/j.chemosphere.2011.02.076 [DOI] [PubMed] [Google Scholar]
- Google My Maps . (2019). UK Pollution hotspots (with distance adjustment). Retrieved May 2, 2019, from https://www.google.com/maps/d/u/0/viewer?mid=1nxsB438uIxBcXx0m8FpfT-3ULszPc6Ka&ll=53.78875006317806%2C-1.7360607608336522&z=14
- GOV.UK . (2018). Ethnicity and type of family or household. Retrieved from https://www.ethnicity-facts-figures.service.gov.uk/ethnicity-in-the-uk/ethnicity-and-type-of-family-or-household
- GOV.UK . (2019). Maternity pay and leave. Retrieved March 15, 2019, from https://www.gov.uk/maternity-pay-leave/leave
- Guxens, M. , Ballester, F. , Espada, M. , Fernández, M. F. , Grimalt, J. O. , Ibarluzea, J. , … Sunyer, J. (2012). Cohort profile: The INMA‐INfancia y Medio Ambiente‐(environment and childhood) project. International Journal of Epidemiology, 41(4), 930–940. 10.1093/ije/dyr054 [DOI] [PubMed] [Google Scholar]
- Hartman, S. , Li, Z. , Nettle, D. , & Belsky, J. (2017). External‐environmental and internal‐health early life predictors of adolescent development. Development and Psychopathology, 29(05), 1839–1849. 10.1017/S0954579417001432 [DOI] [PubMed] [Google Scholar]
- Hawkins, S. S. , Lamb, K. , Cole, T. J. , & Law, C. (2008). Influence of moving to the UK on maternal health behaviours: Prospective cohort study. BMJ, 336(7652), 1052–1055. 10.1136/bmj.39532.688877.25 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Health and Safety Executive . (2019). The Law.
- Heck, K. E. , Braveman, P. , Cubbin, C. , Chávez, G. F. , & Kiely, J. L. (2006). Socioeconomic status and breastfeeding initiation among California mothers. Public Health Reports, 121(1), 51–59. 10.1177/003335490612100111 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Heijkoop, M. (2010). An evolutionary perspective on parental and grandparental investment. Ridderkerk, The Netherlands: Ridderprint Offsetdrukkerij BV. Retrieved from http://igitur-archive.library.uu.nl/dissertations/2010-0204-200158/UUindex.html
- Heinig, M. J. , Follett, J. R. , Ishii, K. D. , Kavanagh‐Prochaska, K. , Cohen, R. , & Panchula, J. (2006). Barriers to compliance with infant‐feeding recommendations among low‐income women. Journal of Human Lactation: Official Journal of International Lactation Consultant Association, 22(1), 27–38. 10.1177/0890334405284333 [DOI] [PubMed] [Google Scholar]
- Huang, R. , & Yang, M. (2015). Paid maternity leave and breastfeeding practice before and after California's implementation of the nation's first paid family leave program. Economics & Human Biology, 16, 45–59. 10.1016/j.ehb.2013.12.009 [DOI] [PubMed] [Google Scholar]
- Ip, S. , Chung, M. , Raman, G. , Chew, P. , Magula, N. , Trikalinos, T. , & Lau, J. (2007). Breastfeeding and maternal and infant health outcomes in developed countries. Evid Technol Asses (Full Rep), 153, 1–186. 10.1542/gr.18-2-15 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Jalaludin, B. , Mannes, T. , Morgan, G. , Lincoln, D. , Sheppeard, V. , & Corbett, S. (2007). Impact of ambient air pollution on gestational age is modified by season in Sydney, Australia. Environmental Health: A Global Access Science Source, 6(16). 10.1186/1476-069X-6-16 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Kaplan, H. S. , & Gangestad, S. W. (2012). Life history theory and evolutionary psychology In Crawford C., & Krebs D. (Eds.), Foundations of evolutionary psychology (pp. 88–113). Psychology Press; Retrieved from. http://psycnet.apa.org/psycinfo/2005-08200-002 [Google Scholar]
- Kelly, Y. J. , Watt, R. G. , & Nazroo, J. Y. (2006). Racial/ethnic differences in breastfeeding initiation and continuation in the United Kingdom and comparison with findings in the United States. Pediatrics, 118(5), e1428–e1435. 10.1542/peds.2006-0714 [DOI] [PubMed] [Google Scholar]
- Kimbro, R. T. (2006). On‐the‐job moms: Work and breastfeeding initiation and duration for a sample of low‐income women. Maternal and Child Health Journal, 10(1), 19–26. 10.1007/s10995-005-0058-7 [DOI] [PubMed] [Google Scholar]
- Klopp, A. , Vehling, L. , Becker, A. B. , Subbarao, P. , Mandhane, P. J. , Turvey, S. E. , … Azad, M. B. (2017). Modes of infant feeding and the risk of childhood asthma: A prospective birth cohort study. Journal of Pediatrics, 190, 192–199.e2. 10.1016/j.jpeds.2017.07.012 [DOI] [PubMed] [Google Scholar]
- Lefebvre, F. (1990). Breast‐feeding among mothers of low birth weight infants. Canadian Family Physician Médecin de Famille Canadien, 36, 1533–1615. Retrieved from. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2280140/pdf/canfamphys00151-0095.pdf [PMC free article] [PubMed] [Google Scholar]
- Lertxundi, A. , Baccini, M. , Lertxundi, N. , Fano, E. , Aranbarri, A. , Martínez, M. D. , … Ibarluzea, J. (2015). Exposure to fine particle matter, nitrogen dioxide and benzene during pregnancy and cognitive and psychomotor developments in children at 15months of age. Environment International, 80, 33–40. 10.1016/j.envint.2015.03.007 [DOI] [PubMed] [Google Scholar]
- Lodge, C. , Tan, D. , Lau, M. , Dai, X. , Tham, R. , Lowe, A. , … Dharmage, S. (2015). Breastfeeding and asthma and allergies: A systematic review and meta‐analysis. Acta Paediatrica, 104, 38–53. 10.1111/apa.13132 [DOI] [PubMed] [Google Scholar]
- Low, B. S. , Hazel, A. , Parker, N. , & Welch, K. B. (2008). Influences on women's reproductive lives: Unexpected ecological underpinnings. Cross‐Cultural Research, 42(3), 201–219. 10.1177/1069397108317669 [DOI] [Google Scholar]
- Luo, Z. C. , Wilkins, R. , & Kramer, M. S. (2006). Effect of neighbourhood income and maternal education on birth outcomes: A population‐based study. CMAJ, 174(10), 1415–1420. 10.1503/cmaj.051096 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Mann, J. (1995). Nurturance or negligence: Maternal psychology and behavioural preference among preterm twins In Barkow J. H., Cosmides L., & Tooby J. (Eds.), The Adapted Mind: Evolutionary Psychology and the Generation of Culture ‐ Google Books. New York: Oxford University Press. [Google Scholar]
- Maternity Action (2014). Breastfeeding in public places. Maternity Action: information sheet. London., 15, 131–135. 10.1177/089033449901500211 [DOI] [Google Scholar]
- Maternity Action . (2018). Continuing to breastfeed when you return to work. Retrieved March 15, 2019, from https://www.maternityaction.org.uk/advice-2/mums-dads-scenarios/6-breastfeeding-rights/continuing-to-breastfeed-when-you-return-to-work/
- Maybury, B. (2016). Report of the Director of Public Health to the meeting of Environment and Waste Overview and Scrutiny Committee to be held on December 20th 2016. Bradford. Retrieved from https://www.gov.uk/government/publications/estimating-local-mortality-burdens-associated-with-particulate‐
- McAndrew, F. , Thompson, J. , Fellows, L. , Large, A. , Speed, M. , & Renfrew, M. J. (2012). Infant Feeding Survey 2010. Leeds: Health and … . Retrieved from http://www.hscic.gov.uk/catalogue/PUB08694/Infant-Feeding-Survey-2010-Consolidated-Report.pdf
- McLennan, D. , Barnes, H. , Noble, M. , Davies, J. , & Garratt, E. (2011). The English indices of deprivation 2010 . Retrieved from https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/6320/1870718.pdf
- Naz, S. , Page, A. , & Agho, K. E. (2016). Household air pollution and under‐five mortality in India (1992–2006). Environmental Health, 15(1), 54 10.1186/s12940-016-0138-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Nettle, D. (2010a). Dying young and living fast: Variation in life history across English neighborhoods. Behavioral Ecology, 21(2), 387–395. 10.1093/beheco/arp202 [DOI] [Google Scholar]
- Nettle, D. (2010b). Why are there social gradients in preventative health behavior? A perspective from behavioral ecology. PLoS ONE, 5(10), e13371 10.1371/journal.pone.0013371 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Nettle, D. (2011). Flexibility in reproductive timing in human females: Integrating ultimate and proximate explanations. Philosophical Transactions of the Royal Society B: Biological Sciences, 366(1563), 357–365. 10.1098/rstb.2010.0073 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Nieuwenhuijsen, M. J. , Dadvand, P. , Grellier, J. , Martinez, D. , & Vrijheid, M. (2013). Environmental risk factors of pregnancy outcomes: A summary of recent meta‐analyses of epidemiological studies. Environmental Health: A Global Access Science Source., 12 10.1186/1476-069X-12-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Office for National Statistics . (2012). Ethnicity and national identity in England and Wales—Office for National Statistics. Retrieved November 8, 2018, from https://www.ons.gov.uk/peoplepopulationandcommunity/culturalidentity/ethnicity/articles/ethnicityandnationalidentityinenglandandwales/2012-12-11
- Office of Environmental Health Hazard Assessment California Environmental Protection Agency . (2010). Public health goal for trihalomethanes in drinking water. Retrieved from https://oehha.ca.gov/media/downloads/water/document/thmphg090910.pdf
- Pearl, M. , Braveman, P. , & Abrams, B. (2001). The relationship of neighborhood sociodemographic characteristics to birthweight among five ethnic groups in California. American Journal of Public Health, 91(11), 1808–1824. 10.2105/AJPH.91.11.1808 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Pedersen, M. , Giorgis‐Allemand, L. , Bernard, C. , Aguilera, I. , Andersen, A. M. N. , Ballester, F. , … Slama, R. (2013). Ambient air pollution and low birthweight: A European cohort study (ESCAPE). The Lancet Respiratory Medicine, 1(9), 695–704. 10.1016/S2213-2600(13)70192-9 [DOI] [PubMed] [Google Scholar]
- Pepper, G. V. , & Nettle, D. (2014). Perceived extrinsic mortality risk and reported effort in looking after health. Human Nature, 25(3), 378–392. 10.1007/s12110-014-9204-5 [DOI] [PubMed] [Google Scholar]
- Poirier, A. , Dodds, L. , Dummer, T. , Rainham, D. , Maguire, B. , & Johnson, M. (2015). Maternal exposure to air pollution and adverse birth outcomes in Halifax, Nova Scotia. Journal of Occupational and Environmental Medicine/American College of . Occupational and Environmental Medicine, 57(12), 1291–1298. 10.1097/JOM.0000000000000604 [DOI] [PubMed] [Google Scholar]
- Quinlan, R. J. (2007). Human parental effort and environmental risk. Proceedings of the Royal Society B: Biological Sciences, 274(1606), 121–125. 10.1098/rspb.2006.3690 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Raisler, J. (2000). Against the odds: Breastfeeding experiences of low income mothers. Journal of Midwifery & Women's Health, 45(3), 253–263. 10.1016/S1526-9523(00)00019-2 [DOI] [PubMed] [Google Scholar]
- Rippeyoung, P. L. F. , & Noonan, M. C. (2012). Is breastfeeding truly cost free? Income consequences of breastfeeding for women. American Sociological Review, 77(2), 244–267. 10.1177/0003122411435477 [DOI] [Google Scholar]
- Rosen‐Carole, C. B. , Auinger, P. , Howard, C. R. , Brownell, E. A. , & Lanphear, B. P. (2017). Low‐level prenatal toxin exposures and breastfeeding duration: A prospective cohort study. Maternal and Child Health Journal, 21(12), 2245–2255. 10.1007/s10995-017-2346-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Ross, E. S. , & Browne, J. V. (2013). Feeding outcomes in preterm infants after discharge from the neonatal intensive care unit (NICU): A systematic review. Newborn and Infant Nursing Reviews, 13(2), 87–93. 10.1053/j.nainr.2013.04.003 [DOI] [Google Scholar]
- Sabatino, G. , Gattoni, M. E. , Castaldi, S. , Piovesan, L. , Tersalvi, C. A. , Auxilia, F. , … Polloni, I. (2015). Air pollution impact on pregnancy outcomes in Como, Italy. Journal of Occupational and Environmental Medicine, 58(1), 47–52. 10.1097/jom.0000000000000630 [DOI] [PubMed] [Google Scholar]
- Schempf, A. , Strobino, D. , & O'Campo, P. (2009). Neighborhood effects on birthweight: An exploration of psychosocial and behavioral pathways in Baltimore, 1995‐1996. Social Science & Medicine, 68(1), 100–110. 10.1016/j.socscimed.2008.10.006 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Scott‐Phillips, T. C. , Dickins, T. E. , & West, S. A. (2011). Evolutionary theory and the ultimate‐proximate distinction in the human behavioral sciences. Perspectives on Psychological Science, 6(1), 38–47. 10.1177/1745691610393528 [DOI] [PubMed] [Google Scholar]
- Shah, P. S. , & Balkhair, T. (2011). Air pollution and birth outcomes: A systematic review. Environment International, 37(2), 498–516. 10.1016/j.envint.2010.10.009 [DOI] [PubMed] [Google Scholar]
- Shenk, M. K. (2011). Our children: Parental decisions—How much to invest in your offspring In Frey U. J., Störmer C., & Willführ K. P. (Eds.), Essential building blocks of human nature (pp. 17–38). Berlin, Heidelberg: Springer Berlin Heidelberg; 10.1007/978-3-642-13968-0 [DOI] [Google Scholar]
- Shoji, H. , & Koletzko, B. (2007). Oxidative stress and antioxidant protection in the perinatal period. Current Opinion in Clinical Nutrition and Metabolic Care, 10(3), 324–328. 10.1097/MCO.0b013e3280a94f6d [DOI] [PubMed] [Google Scholar]
- Silvers, K. M. , Frampton, C. M. , Wickens, K. , Pattemore, P. K. , Ingham, T. , Fishwick, D. , … Epton, M. J. (2012). Breastfeeding protects against current asthma up to 6 years of age. Journal of Pediatrics, 160(6), 991–996.e1. 10.1016/j.jpeds.2011.11.055 [DOI] [PubMed] [Google Scholar]
- Smith, R. B. , Edwards, S. C. , Best, N. , Wright, J. , Nieuwenhuijsen, M. J. , & Toledano, M. B. (2016). Birth weight, ethnicity, and exposure to trihalomethanes and haloacetic acids in drinking water during pregnancy in the Born in Bradford cohort. Environmental Health Perspectives, 124(5), 681–688. 10.1289/ehp.1409480 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Sonnenschein‐van Der Voort, A. M. M. , Jaddoe, V. W. V. , Van Der Valk, R. J. P. , Willemsen, S. P. , Hofman, A. , Moll, H. A. , … Duijts, L. (2012). Duration and exclusiveness of breastfeeding and childhood asthma‐related symptoms. European Respiratory Journal, 39(1), 81–89. 10.1183/09031936.00178110 [DOI] [PubMed] [Google Scholar]
- Stanfield, K. M. , Wells, J. C. , Fewtrell, M. S. , Frost, C. , & Leon, D. A. (2012). Differences in body composition between infants of South Asian and European ancestry: The London mother and baby study. International Journal of Epidemiology, 41(5), 1409–1418. 10.1093/ije/dys139 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Stefanidou, M. , Maravelias, C. , & Spiliopoulou, C. (2009). Human exposure to endocrine disruptors and breast milk. Endocrine, Metabolic & Immune Disorders Drug Targets, 9(3), 269–276. 10.2174/187153009789044374 [DOI] [PubMed] [Google Scholar]
- Tischer, C. G. , Hohmann, C. , Thiering, E. , Herbarth, O. , Müller, A. , Henderson, J. , … Heinrich, J. (2011). Meta‐analysis of mould and dampness exposure on asthma and allergy in eight European birth cohorts: An ENRIECO initiative. Allergy: European Journal of Allergy and Clinical Immunology, 66(12), 1570–1579. 10.1111/j.1398-9995.2011.02712.x [DOI] [PubMed] [Google Scholar]
- Tommy's . (2019). Feeding your premature baby. Retrieved February 3, 2019, from https://www.tommys.org/pregnancy-information/pregnancy-complications/premature-birth/your-babys-time-hospital/feeding-your-premature-baby
- Tully, K. P. , & Ball, H. L. (2013). Trade‐offs underlying maternal breastfeeding decisions: A conceptual model. Maternal & Child Nutrition, 9(1), 90–98. 10.1111/j.1740-8709.2011.00378.x [DOI] [PMC free article] [PubMed] [Google Scholar]
- UCSF Children's Hospital . (2004). Very low and extremely low birthweight infants. Retrieved from https://www.ucsfbenioffchildrens.org/pdf/manuals/20_VLBW_ELBW.pdf
- UNICEF . (2002). Facts for life . United Nations Children's fund. Retrieved from https://books.google.co.uk/books?hl=en&lr=&id=GAFgWda-2NMC&oi=fnd&pg=PR18&dq=UNICEF+2002+facts+for+life+3rd+edition&ots=QjutsDkvAP&sig=VWAV8bgmOR8JXhnoQHsHnZ0TsOw#v=onepage&q=breast&f=false
- Victora, C. G. , Bahl, R. , Barros, A. J. D. , França, G. V. A. , Horton, S. , Krasevec, J. , … Rollins, N. C. (2016). Breastfeeding in the 21st century: Epidemiology, mechanisms, and lifelong effect. The Lancet, 387(10017), 475–490. 10.1016/S0140-6736(15)01024-7 [DOI] [PubMed] [Google Scholar]
- Virgo, S. , & Sear, R. (2016). Area‐level mortality and morbidity predict abortion proportion in England and Wales. Evolution and Human Behavior, 37(5), 1–31. 10.1016/j.evolhumbehav.2016.03.001 26766895 [DOI] [Google Scholar]
- Voland, E. (1998). Evolutionary ecology of human reproduction. Annual Review of Anthropology, 27, 347–374. 10.1146/annurev.anthro.27.1.347 [DOI] [PubMed] [Google Scholar]
- Waller, K. , Swan, S. H. , Delorenze, G. , & Hopkins, B. (1998). Trihalomethanes in drinking water and spontaneous abortion. Epidemiology, 9(2), 134–140. Retrieved from. http://www.ncbi.nlm.nih.gov/pubmed/26641824, 10.1097/00001648-199803000-00006 [DOI] [PubMed] [Google Scholar]
- Wambach, K. , & Riordan, J. (2016). In Wambach K., & Riordan J. (Eds.), (Enhanced f)Breastfeeding and human lactation. Burlington: Jones & Bartlett Learning. [Google Scholar]
- Waynforth, D. (2015). Reduced birth intervals following the birth of children with long‐term illness: Evidence supporting a conditional evolved response. Biology Letters, 11(10), 20150728 10.1098/rsbl.2015.0728 [DOI] [PMC free article] [PubMed] [Google Scholar]
- Wells, J. C. K. , Nesse, R. M. , Sear, R. , Johnstone, R. A. , & Stearns, S. C. (2017). Evolutionary public health: Introducing the concept. The Lancet, 390(10093), 500–509. 10.1016/S0140-6736(17)30572-X [DOI] [PubMed] [Google Scholar]
- Williamson, I. R. , & Sacranie, S. M. (2012). Nourishing body and spirit: Exploring British Muslim mothers' constructions and experiences of breastfeeding. Diversity and Equality in Health and Care, 9, 113–123. Retrieved from. http://diversityhealthcare.imedpub.com/nourishing-body-and-spirit-exploring-british-muslim-mothers-constructions-and-experiences-of-breastfeeding.pdf [Google Scholar]
- World Health Organization . (2005). Trihalomethanes in drinking‐water. Background document for development of WHO guidelines for drinking‐water quality. Retrieved from http://www.who.int/water_sanitation_health/dwq/chemicals/THM200605.pdf
- Wright, J. , Small, N. , Raynor, P. , Tuffnell, D. , Bhopal, R. , Cameron, N. , … West, J. (2013). Cohort profile: The born in bradford multi‐ethnic family cohort study. International Journal of Epidemiology, 42(4), 978–991. 10.1093/ije/dys112 [DOI] [PubMed] [Google Scholar]
- Zaidi, F. (2014). Challenges and practices in infant feeding in Islam. British Journal of Midwifery, 22(3), 167–172. 10.12968/bjom.2014.22.3.167 [DOI] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data S1 Supporting information