Abstract
The aim of our study was to examine whether pre-natal exposure to dichlorodiphenyldichloroethylene (DDE) increases the risk of lower respiratory tract infections (LRTIs) and wheeze in infants.
The study is based on a birth cohort of 1,455 mother–child pairs. Maternal serum concentrations of DDE, polychlorinated biphenyls (PCBs) and hexachlorobenzene (HCB) were measured during pregnancy. Parental reports on LRTI and wheeze were obtained when children were 12–14 months old.
35.4% of children developed at least one LRTI episode and 33.6% at least one wheezing episode during their first 12–14 months of life. Median DDE, PCBs and HCB concentrations were 116.3, 113.7 and 46.4 ng·g−1 lipid, respectively. DDE concentrations were associated with LRTI risk (relative risk (RR) per 10% increase 1.11, 95% CI 1.00–1.22), also after adjustment for PCBs and HCB. In all quartiles of DDE exposure, the risk of LRTI was increased compared with the lowest quartile, but the increase was statistically significant only in the third quartile (RR 1.33, 95% CI 1.08–1.62). No association was observed for PCBs and HCB. Results were similar for wheeze.
This study suggests that pre-natal DDE exposure is associated with a higher risk of LRTI and wheeze in infants independently of exposure to other organochlorine compounds.
- Children
- dichlorodiphenyldichloroethylene
- lower respiratory tract infections
- organochlorine compounds
- pre-natal exposure
- wheeze
Acute respiratory infections (ARIs) are a worldwide cause of morbidity and mortality in children aged <5 yrs [1]. Lower respiratory tract infections (LRTIs), mainly pneumonia and bronchiolitis, are considered to be the major components that account for the global burden of disease from ARI among young children. Several risk factors have been reported to increase vulnerability to LRTI during infancy and childhood, such as tobacco exposure, type and duration of breastfeeding, and familiar history of atopy or allergic asthma [2, 3]. Moreover, growing evidence suggests that pre-natal exposure to organochlorine compounds (OCs), mainly polychlorinated biphenyl (PCB)-153 and dichlorodiphenyldichloroethylene (DDE), may increase the risk of respiratory symptoms during the first years of life, even at low exposure levels [4–9]. In addition, LRTIs are one of the major risk factors for developing asthma later in life [2, 10] and pre-natal DDE exposure has also been associated with asthma and wheezing in children aged 4 [11] and 6 yrs [12].
OCs are synthetic persistent organic pollutants used worldwide and distributed throughout the environment, food and human tissues. Immunological effects of OCs have been reported in studies conducted both in animals [13–16] and humans [17–20]. However, previous epidemiological studies on the association between OCs and LRTI have not been able to clearly determine which compound (PCBs, DDE or other OCs) was responsible for these effects due to the high correlation between concentrations of individual compounds [5, 6, 8]. In a birth cohort study in Sabadell, Spain, Sunyer et al. [9] were the first to indentify DDE as the main responsible compound, but the study was too small to draw strong conclusions or to examine the role of other risk factors as possible effect modifiers.
In our study, we use a larger Spanish birth cohort, including the previous study [9] in Sabadell, Spain to: 1) provide more precise estimates for the effect of pre-natal DDE exposure on occurrence of LRTI and wheeze in infants; 2) isolate these effects from those of other OCs, including hexachlorobenzene (HCB) and PCBs; and 3) explore the role of other risk factors in this association, including maternal smoking, maternal history of atopy and allergic asthma, and breastfeeding practices. In addition, since recent studies suggest that a high level of adherence to the Mediterranean diet during pregnancy protects against the development of asthma and atopy in children [21], maternal diet during pregnancy is also explored as a possible effect modifier.
METHODS
Study population
This study is based in three Spanish regions (Gipuzkoa, Sabadell and Valencia) belonging to the INMA (Infancia y Medio Ambiente (Environment and Childhood)) Project in Spain [22]. All regions followed the same protocol and started recruiting pregnant females into the cohort between 2004 and 2008 (Sabadell, n=657; Valencia, n=855; and Gipuzkoa, n=638). Pregnant females coming to the first trimester routine antenatal care visit in the main public hospital or health centre of reference and who fulfilled the inclusion criteria (aged >16 yrs, intention to deliver in the city and no problems of communication) were recruited. Protocol details are described elsewhere [22]. This study was conducted with the approval of the hospital ethics committees in the participating regions and written informed consent was obtained from the parents of all children.
Outcomes
Information about physician-confirmed diagnosis of LRTI was obtained from parents through questionnaires when children were 1 yr old (mean±sd in Valencia was 12.4±1.1 months, in Gipuzkoa was 14.3±1.2 months and in Sabadell was 14.5±0.7 months). Occurrence of an LRTI episode was defined as a positive answer to both a general question (“Since the last interview, has the doctor told you that your child has had a chest infection?”) and a specific question on the type of infection (bronchiolitis, bronchitis or pneumonia) determined by the doctor. Children with negative answers to both questions were defined as not having LRTI, and those reporting positive answers to both questions were defined as having LRTI. Those whose answers to both questions did not match (n=58) were excluded from the study. Wheezing was defined as a positive answer to the question “Has your child ever experienced whistling or wheezing from the chest, but not noisy breathing from the nose from birth to 12–14 months of age?”. All these questions were based on the validated International Study of Asthma and Allergies in Childhood (ISAAC) questionnaire [23].
Exposure assessment
Concentrations of OCs (HCB, p,p′-DDE, and PCB congeners 28, 118, 138, 153 and 180) in maternal serum extracted between the 7th week and the 26th week of pregnancy (median 12.9 weeks) from peripheral veins, were stored in crystal tubes at 20°C and analysed with a gas chromatograph using methods described elsewhere [24]. The limits of detection (LOD) were 0.071 ng·mL−1 in Sabadell and Gipuzkoa and between 0.01 and 0.071 ng·mL−1 in Valencia. International intercalibration exercises showed that differences of levels between regions were not due to laboratory differences. For comparison purposes, values in Valencia <0.071 ng·mL−1 were set as nondetectable. Samples with nondetectable levels were then set at a value of half the LOD. The sum of PCBs (ΣPCBs) was calculated by summing the concentrations of all individual congeners except PCB-28, which was detectable in <1% of samples. PCB-138, -153 and -180 were the predominant congeners. All exposures are expressed on a lipid basis in ng·g−1 of lipid using the method described elsewhere [25]. Correlations between lipid adjusted and not adjusted values were high (0.97 for p,p′-DDE and 0.95 for ΣPCBs).
Other variables
Information on covariables was extracted from the questionnaires answered by the females during the 3rd trimester of pregnancy and at child age 12–14 months. Covariables of interest for the current study included: maternal age, social class (based on the International Standard Classification of Occupations), education and country of origin of the mother, maternal smoking during pregnancy, maternal smoking during the year after birth, parity (first child or not), daycare attendance, duration of predominant breastfeeding (never breastfeeding, breastfeeding 1–16 weeks, 17–24 weeks, >24 weeks), maternal history of atopy and/or allergic asthma, and maternal consumption of meat, fish and vegetables during pregnancy (divided into tertiles). As maternal atopy and allergic asthma were highly correlated (p<0.001), we combined them into a new single variable: “atopic–asthmatic mother”. Pre-pregnancy weight of the mother, gestational age and weight at birth were collected from clinical records or reported by mothers.
Statistical methods
Out of the initial population of recruited mother–child pairs (n=2,150), 279 were lost to follow-up at the time of the age 12–14-month visit and 416 had missing information regarding exposure to OCs, regarding one of the outcomes or regarding country of origin, resulting in 1,455 mother–child pairs with complete exposure–outcome information. Because of their very different exposure profiles, analyses were performed separately for Spanish mothers (n=1,342) and Latin-American mothers (n=79); mothers of other origin were not included (n=34).
Some of the covariables of interest for our analysis had missing information (between 0.1% and 3.4%). These missing values were imputed by multiple imputation [26]. This method is based on conditioning the missing variables’ density to given predictor variables, which in our case were country of origin, parity, gestational age, maternal age, maternal pre-pregnancy weight, maternal social class and maternal education, sex and birthweight, duration of predominant breastfeeding, daycare attendance, smoking during pregnancy or 1 yr after birth, being an atopic–asthmatic mother, maternal consumption of meat, fish and vegetables during pregnancy, and lipid and OC levels in maternal blood. These imputations were performed separately by region of study.
A log-binomial regression model was used to analyse the relationship between concentrations of DDE, HCB and ΣPCBs with LRTI and wheezing. Generalised additive models were used to graphically examine the shape of relationships between OC exposure and outcome variables. These did not show statistically significant evidence for a departure from (log)linear relationships (p-values for gain in linearity were between 0.12 and 0.23) (fig. 1). However, since the evidence was graphically not very strong, especially for PCBs, we performed analyses using OC concentrations both as (log-transformed) continuous exposure variables and as exposure categories using quartiles as cut-offs. Potential confounder variables were included one by one in the model. Variables were retained in the final model if they were related to the outcome (p<0.2), or changed the β-coefficient for the relationship between exposure and outcome by >10%. Variables that did not meet these criteria, but which were considered important risk factors for LRTI or wheezing were also included in the final model (maternal smoking during pregnancy, maternal smoking during the year after birth, social class and duration of predominant breastfeeding). Covariables included in the final models for each outcome are indicated in the results section.
Generalised additive models to examine the shape of relationships between exposure to organochlorine compounds and lower respiratory tract infections (LRTIs): a) dichlorodiphenyldichloroethylene (DDE) (p-value for gain in linearity=0.23), b) hexachlorobenzene (HCB) (p=0.12), and c) sum of polychlorinated biphenyls (ΣPCBs) (p=0.13).
The influence of multi-pollutants on the relationship between DDE, HCB or ΣPCBs and LRTI was examined by including these compounds together in one model. Given that some characteristics of the participants and the mixture of OCs varied by region, sensitivity analyses were carried out stratifying by region. Analyses with the one pollutant model were further stratified by potential effect modifiers, such as atopic–asthmatic mother, maternal smoking during pregnancy, maternal smoking 1 yr after pregnancy, duration of predominant breastfeeding, consumption of vegetables and fruit, and fish consumption during pregnancy. Wald tests were used to test the statistical significance of interaction terms. Since a similar analysis of the Sabadell cohort has been published [9], we performed a sensitivity analysis excluding subjects from Sabadell.
Separate analyses for the Latin-American population followed the same methodology as those for the Spanish population. Tertiles of DDE concentrations were created instead of quartiles, because of the small population. HCB and the ΣPCBs were not analysed due to the very low concentrations detected within this population (below the LOD in 54.4% and 78.5% of samples). All analyses were conducted using STATA 10 (StataCorp, College Station, TX, USA).
RESULTS
Spanish population: main analysis
There were significant differences between Spanish mother–child pairs included in the main analyses and those excluded (table S1); mothers included were older, had a higher education level, a higher pre-pregnancy weight and smoked less. They also ate more fruit, vegetables and fish, and breastfed their children for longer. Among included subjects, there were fewer pre-term and low birthweight children and a higher percentage of wheezing cases. Maternal pre-natal concentrations of HCB and ΣPCBs were significantly higher among included participants, but DDE concentrations and prevalence of LRTI did not significantly differ from excluded subjects (p=0.21 and p=0.32, respectively).
A total of 35.4% and 33.6% of the children had at least one episode of LRTI or wheezing, respectively, during their first year of life (table 1). LRTI and wheeze were highly correlated, with 76% of children with LRTI also reporting wheezing symptoms. Males, children attending daycare, those with maternal atopic asthma and smoking, those with multiparous mothers and those who were breastfed for a shorter period of time had a higher risk of ever having LRTI and/or wheezing symptoms (table 1).
Maternal levels of DDE were higher among pre-term, low-birthweight children and those not attending daycare services (table 2). Mothers with higher levels of DDE were older, had a higher pre-pregnancy weight and had lower education levels. A higher consumption of meat was also related to higher maternal DDE levels. In general, concentrations of HCB were much lower (46.4 ng·g−1 of lipid) than DDE (116.3 ng·g−1 of lipid) or ΣPCBs (113.7 ng·g−1 of lipid) (table 3). Correlation coefficients were 0.43 (DDE and ΣPCBs), 0.49 (DDE and HCB) and 0.40 (ΣPCBs and HCB) (all p<0.001).
Associations between the covariables in the final regression models and the outcomes are presented in table S2. Risk of LRTI increased with increasing DDE exposure and was statistically significant after adjustment for potential confounders (crude relative risk (RR) for 10% increase in DDE concentration 1.04, 95% CI 0.95–1.14 and adjusted RR 1.11, 95% CI 1.00–1.22). Adjustment for other pollutants gave a RR of borderline statistical significance (RR 1.11, 95% CI 0.99–1.24). In all quartiles of DDE, the risk of LRTI was increased compared with the lowest quartile, but the increase was statistically significant only in the 3rd quartile (adjusted RR 1.33, 95% CI 1.08–1.62). This increase remained after adjustment for other OCs (RR for 3rd quartile DDE 1.40, 95% CI 1.13–1.73) (table 4). Pre-natal HCB levels did not increase the RR of LRTI. In the multi-pollutant model, the risk of LRTI in the highest quartile of PCB exposure was statistically significantly lower than in the lowest quartile (table 4). Risk estimates for wheezing were very similar to those found for LRTI (table 5).
The association between DDE and LRTI did not differ between strata defined by region (table S3), duration of predominant breastfeeding, maternal smoking during pregnancy or the first year of life, atopic–asthmatic mother or by maternal consumption of vegetables and fruit, meat or fish (data not shown). Sensitivity analysis adjusting for pre-term births or excluding these children (n=51) from the model did not modify the results. Analysis without imputed data also provided similar results.
Latin-American population
Among mothers of Latin-American origin, DDE was detected in all samples, with a median concentration of 385.0 ng·g−1 of lipid (table 3). The risk estimate for continuous DDE exposure in Latin-American mothers was very similar to those in Spanish mothers, but did not reach statistically significance (adjusted RR 1.14, 95% CI 0.92–1.42). When analysing by tertiles, estimates became statistically significant after adjustment (second tertile of DDE RR 2.59, 95% CI 1.00–6.66; and the third tertile RR 2.89, 95% CI 1.10–7.55). Results were similar for wheezing (table 6).
DISCUSSION
The present study suggests that pre-natal exposure to DDE is associated with a higher risk of LRTI and wheeze in infants. The DDE effect was independent of HCB or PCB exposure and was not clearly modified by other risk factors, including maternal smoking, maternal medical history of atopy/asthma, maternal dietary habits in pregnancy or breastfeeding practices. These results confirm our previous findings in a subsample of the present study (using a different definition of the outcome) [9].
Only three other cohort studies, in Canada [5, 6], Menorca (Spain) [11, 12] and Sweden [9], assessed respiratory infections or wheezing in young children in relation to pre-natal DDE exposure, and results are inconsistent. In the Canadian cohort [5, 6], with higher levels of DDE (geometric mean 294 ng·g−1 of lipid), there was no clear increase in risk of LRTI with DDE concentrations at age 6–12 months [5], but at age 5 yrs, exposure to OCs (DDE and others assessed using PCB-153 concentrations as surrogate) did increase risk of LRTI [6]. The same study did find increased risks of upper respiratory tract infections and/or otitis in both age groups. The Swedish study [9] found that DDE exposure was related to a nonstatistically significant decrease in LRTI risk, but DDE levels were lower than those in our study (median 88 ng·g−1 of lipid) and children were assessed at age 3 months, providing little time to develop infections [8]. In the Menorcan birth cohort [11, 12], with pre-natal DDE median levels of ∼170 ng·g−1 of lipid, DDE was associated with a higher risk of wheeze and asthma at ages 4 and 6.5 yrs, but not to earlier wheezing or LRTI during the first year of life [11, 12]. This somewhat conflicts with our results, but the number of children in the Menorca cohort (n<400) may have been too small to detect early effects. It will be important to assess the DDE effects in our present cohort at older ages. Most of the previous studies could not separate the effects of DDE on LRTI from those of other OCs [5, 6, 8], whereas our study clearly identifies DDE as the main responsible compound. This could be partly explained by the different correlations between DDE and other OCs across studies; in the Swedish [8] and Canadian cohorts [5, 6], correlations between DDE and other OCs were between 0.66 and 0.89, whereas correlations among the nonmigrant population of the present Spanish cohort were <0.49. In our study, risk of LRTI and wheezing decreased in the highest quartile of PCBs exposure; this decrease was statistically significant only after adjustment for DDE and HCB. We do not have a real explanation for these results, but this may be a chance result, unexplained by confounding or a problem of multicollinearity between OCs. Our quartile results (with significant increase only in the third quartile) indicate that the association between DDE and LRTIs may not be strictly monotonic, even though we observed a linear trend with continuous exposure, and additional linear spline analyses showed no statistical evidence for differences between spline slopes (table S4). Nonmonotonic functions have also been reported by others [5, 6].
In our cohort, mothers from Latin-American origin showed very different patterns of OCs exposure from Spanish mothers, with very high DDE levels, low levels of PCBs and HCB, and low correlations between DDE and other OCs. Diet, other lifestyle factors and differences in industrial development may explain part of these differences [27], together with the fact that in Latin-America the use of dichlorodiphenyltrichloroethane (DDT), the parent compound of DDE, lasted until the late 1990s for agricultural and malaria vector control purposes [28]. The similarity of our DDE effect in both Spanish and Latin-American children indicates that this effect may apply widely to populations with different OCs exposure patterns. This is of especial interest, as nowadays the practice of using DDT for malaria vector control is still present or planned to be introduced in many developing countries with endemic malaria [29]. However, our results are limited by the small size of our Latin-American population, and further studies in areas with high levels of DDE and low levels of other OCs are indicated.
The mechanisms by which DDE and other OCs may produce LRTI and wheeze are not fully understood. However, some studies have shown an association between DDE exposure levels and the uncontrolled production of cytokines and the increase of nitric oxide production in macrophages, contributing to inflammatory reactions, cytokine imbalance and immune dysregulation [16, 17, 20]. DDE has also been associated with altered levels and a reduced viability and proliferation capacity of immune system cells (macrophages, lymphocytes and monocytes) [8, 17, 18], mainly through apoptosis (programmed cell death) [13, 30], which seems to be caused by oxidative stress [30]. Although apoptosis plays a very important role under normal physiological conditions, when not regulated, apoptosis can contribute to immune dysregulation and immunodeficiency [31]. Moreover, recent studies suggest that apoptotic cells actively regulate the immune response by releasing immunosuppressive cytokines (e.g. transforming growth factor-β1) and by suppressing the secretion of pro-inflammatory cytokines (e.g. tumour necrosis factor-α), indicating an immunosuppressive response [30, 32] that could lead to an increased risk of contracting infections. Further studies are needed to understand better the mechanisms by which DDE interferes with the immune system [31]; although cell counts serve as a general indicator of immune status, future research should focus on the performance of cytokine assays, as they can provide a more mechanistic examination of the effect of exposure [32].
A limitation of our study is the lack of serology or culture to confirm the LRTI diagnosis. However, we used repeated questionnaire items to define LRTI and our results were consistent, whichever definition criteria was selected (data not shown). The fact that we found similar results for wheezing, a related outcome for LRTI at this age, and that covariables in the final model were associated with the outcomes as expected, also provide strength to our results. Our cohort is somewhat selective, as loss to follow-up and incompleteness of the questionnaires occurred more often in younger mothers with lower educational levels [27]; this is also reflected in the differences between our included study population and the excluded mothers. It is unlikely that this has led to spurious associations, but it means that these groups of the population are under-represented in our sample. We observed somewhat inconsistent results between the regions of our cohort (table S3), but there was no evidence of heterogeneity between regions (p-value for interaction 0.30–0.82) and inconsistencies may have resulted from small comparison groups, especially in the quartile analyses. Meta-analyses of the estimates of each region were performed and results were similar to those of the pooled analysis (table S4). Any small differences between regions might be due to the percentage of mothers reporting LRTI or wheezing during the first year of life of their children, which was lower in Valencia than in Gipuzkoa or Sabadell. This is probably because, in Valencia, respiratory questionnaires were administered at ∼12 months of age, instead of at 14 months in the other regions. Strengths of the current study are its prospective study design and large population size. Also, in a sensitivity analysis, we were able to show that the DDE effect was not only due to the influence of the one region for which data had already been published (data not shown) [9].
Conclusions
The present study reinforces the hypothesis that pre-natal exposure to DDE is associated with a higher risk of LRTI and wheeze in infants. As LRTIs cause substantial morbidity in infancy, and LRTI and wheeze are possible risk factors for subsequent childhood asthma, particular attention should be paid to these in countries where DDT is currently used for malaria control.
Acknowledgments
The authors would like to thank the families participating in this study. They are also grateful to all fieldworkers for their assistance in contacting the families and administering the questionnaires. A full listing of the INMA project researchers can be found at www.proyectoinma.org.
Footnotes
This article has supplementary material available from www.erj.ersjournals.com
Earn CME accreditation by answering questions about this article. You will find these at the back of the printed copy of this issue or online at www.erj.ersjournals.com/misc/cmeinfo.xhtml
Support Statement
This study was supported by the Instituto de Salud Carlos III, Red de Grupos INMA (G03/176; CB06/02/0041). It has also been partially supported by the Fundació La Caixa (00/077-00), the Instituto de Salud Carlos III, Red de Centros RCESP (C03/09), FIS (03/1615, 04/1436, 04/1509, 04/1112, 04/1931, 05/1079, 05/1052, 06/0867, 06/1213, 07/0314, 08/1151, 09/02647), the Generalitat de Catalunya (CIRIT 1999SGR00241), the Diputación Foral de Gipuzkoa (DFG06/004), the Dept of Health of the Basque Government (2005111093), the Conselleria de Sanitat Generalitat Valenciana and the Fundació Roger Torné.
Statement of Interest
None declared.
- Received January 21, 2011.
- Accepted October 13, 2011.
- ©ERS 2012
REFERENCES
