Neurodevelopmental disorders are relatively common and pose a substantial challenge to society (Froehlich et al., Reference Froehlich, Lanphear, Epstein, Barbaresi, Katusic and Kahn2007; Jarbrink et al., Reference Jarbrink, Fombonne and Knapp2003; Kogan et al., Reference Kogan, Strickland, Blumberg, Singh, Perrin and van Dyck2008; Newton, Reference Newton2012). For some conditions the diagnosis rates have increased, but the reasons behind these apparent time trends remain largely unknown. Improved awareness and widened diagnostic criteria may contribute, such as for attention deficit/hyperactivity disorder (ADHD) and autism spectrum disorders (ASD), but probably do not explain the whole increase. Both ADHD and ASD are childhood-onset chronic conditions of moderate to high heritability (Anckarsater et al., Reference Anckarsater, Lundstrom, Kollberg, Kerekes, Palm, Carlstrom and Lichtenstein2011; Martin et al., Reference Martin, Scourfield and McGuffin2002; Parr et al., Reference Parr, Le Couteur, Baird, Rutter, Pickles and Fombonne2011). However, their precise etiologies remain enigmatic, and the role of environmental factors acting as triggers or contributors to general vulnerability should not be disregarded (Sandin et al., Reference Sandin, Lichtenstein, Kuja-Halkola, Larsson, Hultman and Reichenberg2014).
Epidemiological and experimental studies indicate that exposure to air pollution from road traffic may induce systemic inflammation and increase the risk of several diseases related to inflammation, such as asthma, allergy, and cardiovascular diseases (Mills et al., Reference Mills, Donaldson, Hadoke, Boon, MacNee, Cassee and Newby2009; Nordling et al., Reference Nordling, Berglind, Melen, Emenius, Hallberg, Nyberg and Bellander2008; Panasevich et al., Reference Panasevich, Leander, Rosenlund, Ljungman, Bellander, de Faire and Nyberg2009). Systemic inflammation can also contribute to neuronal injury and affect the development of central nervous system (Hagberg & Mallard, Reference Hagberg and Mallard2005). Recent epidemiological studies have shown associations between exposure to air pollution from road traffic or other sources and adverse neurodevelopmental effects in children (Becerra et al., Reference Becerra, Wilhelm, Olsen, Cockburn and Ritz2013; Calderon-Garciduenas et al., Reference Calderon-Garciduenas, Engle, Mora-Tiscareno, Styner, Gomez-Garza, Zhu and D’Angiulli2011; Dix-Cooper et al., Reference Dix-Cooper, Eskenazi, Romero, Balmes and Smith2012; Guxens et al., Reference Guxens, Aguilera, Ballester, Estarlich, Fernandez-Somoano and Lertxundi2012; Jung et al., Reference Jung, Lin and Hwang2013; Morales et al., Reference Morales, Julvez, Torrent, de Cid, Guxens, Bustamante and Sunyer2009; Siddique et al., Reference Siddique, Banerjee, Ray and Lahiri2011; Volk et al., Reference Volk, Hertz-Picciotto, Delwiche, Lurmann and McConnell2011, Reference Volk, Lurmann, Penfold, Hertz-Picciotto and McConnell2013; Vrijheid et al., Reference Vrijheid, Martinez, Aguilera, Bustamante, Ballester, Estarlich and Project2012; Windham et al., Reference Windham, Zhang, Gunier, Croen and Grether2006). However, more studies are needed to assess causality, particularly since the association may be confounded by socio-economic and socio-demographic characteristics (Bhasin & Schendel, Reference Bhasin and Schendel2007; Flouri et al., Reference Flouri, Mavroveli and Tzavidis2012). Furthermore, it is not known whether there are specific periods of increased vulnerability.
The primary objective of this study was to investigate the relation between exposure to air pollution from road traffic and the risk of neurodevelopmental disorders in children, especially ASD and ADHD. In particular, the influence of exposure during potentially important time windows, such as the fetal and infancy periods, was in focus.
Materials and Methods
Study Population
Children from the Child and Adolescent Twin Study in Sweden (CATSS), an ongoing longitudinal cohort study that targets all twins born in Sweden since July 1, 1992, were the participants (Anckarsater et al., Reference Anckarsater, Lundstrom, Kollberg, Kerekes, Palm, Carlstrom and Lichtenstein2011). In this project, the twins born during 1992–2000 were included. Parents of 17,220 9-year-old twins were contacted and interviewed about their children's somatic and mental health as well as social environment (Figure 1). During the first 3 years of the study, 12-year-old twins were also included. Since the air pollution exposure assessment methodology was restricted to Stockholm County, 4,980 twins born in this area were selected and 3,426 completed neurodevelopmental assessment (response rate: 68.8%). The study was approved by the Regional Ethical Review Board in Stockholm, Sweden.
FIGURE 1 Summary of participation and response rates.
Health Outcome Assessment
Children's neurodevelopmental outcomes were measured using the Autism-Tics, ADHD, and other Comorbidities inventory (A-TAC) telephone interviews developed at the Institute of Neuroscience and Physiology, Child and Adolescent Psychiatry, Gothenburg University (Hansson et al., Reference Hansson, Svanstrom Rojvall, Rastam, Gillberg, Gillberg and Anckarsater2005). The A-TAC comprises 178 symptom questions from a lifetime perspective and is designed as an open-access and comprehensive tool for screening childhood ASD and other targeted disorders based on the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) criteria. Response options for each question were coded as 0 for ‘No,’ 0.5 for ‘Yes,’ to some extent, and 1.0 for ‘Yes’. In two previous validation studies, autistic-like traits were assessed by the sum scores of 12 items (based on the DSM-IV criteria) or 17 items (by adding one additional item from the flexibility and two additional items each from the language and social interaction modules; Larson et al., Reference Larson, Anckarsater, Gillberg, Stahlberg, Carlstrom, Kadesjo and Gillberg2010; Hansson et al., Reference Hansson, Svanstrom Rojvall, Rastam, Gillberg, Gillberg and Anckarsater2005). In order to comprise the primary symptoms of ADHD, scores of 18 (based on the DSM-IV criteria) or 19 items (by adding one additional item from the impulsivity module) were summed. Cut-off values for the sum scores with high sensitivity and specificity from previous validation studies were used in the current study to resemble the probabilities of clinical diagnoses and severity of both diseases: ASD ≥ 4.5 for DSM-IV criteria and for the lower cut-off value of extended diagnostic criteria, ASD ≥ 8.5 for the higher cut-off value of extended diagnostic criteria, ADHD ≥ 8 for DSM-IV criteria, and ADHD ≥ 6 for the lower and ADHD ≥ 12.5 for the higher cut-off values of extended diagnostic criteria. Detailed information on the psychometric properties of the A-TAC is provided elsewhere (Anckarsater et al., Reference Anckarsater, Lundstrom, Kollberg, Kerekes, Palm, Carlstrom and Lichtenstein2011; Hansson et al., Reference Hansson, Svanstrom Rojvall, Rastam, Gillberg, Gillberg and Anckarsater2005; Larson et al., Reference Larson, Anckarsater, Gillberg, Stahlberg, Carlstrom, Kadesjo and Gillberg2010).
Exposure Assessment
The air pollution concentrations at residential addresses during mother's pregnancy, child's first year of life, and the year before the neurodevelopmental assessment were estimated by dispersion models, described in detail elsewhere (Bellander et al., Reference Bellander, Berglind, Gustavsson, Jonson, Nyberg, Pershagen and Jarup2001; Gruzieva et al., Reference Gruzieva, Bellander, Eneroth, Kull, Melen, Nordling and Pershagen2012). Briefly, the residential history of the study participants was obtained from taxation authorities and geo-coded using a property register maintained by the Swedish mapping, cadastral, and land registration authority. The address information was linked with historical emission databases to obtain annual average levels of nitrogen oxides (NOx) and particulate matter (PM) with less than 10 μm of diameter (PM10). Residence time-weighted NOx and PM10 concentrations related to road traffic emissions were calculated for each trimester and over the mother's pregnancy period, the child's first year, and the ninth year of life. Furthermore, daily 24-hour mean NOx and PM10 levels from suburban stations were used to calculate the NOx and PM10 levels during each trimester of pregnancy, which were taken into account in sensitivity analyses. Imputation for missing values of NOx and PM10 in the trimester-specific analyses was performed using predictions from rooftop measurements of both pollutants from a monitoring station in the center of Stockholm.
Other Covariates
Information on gender (male/female), parity (first/second/third/fourth, or later), gestational age (<37 weeks/≥37 weeks), birth weight (<2,500 g/≥2,500 g), maternal age at birth (<25/25–29/30–34/≥35 years old), maternal smoking during pregnancy (no cigarette/1–9 cigarettes per day/≥10 cigarettes per day) was obtained from the Medical Birth Register (National Board of Health and Welfare, 2003). Using the longitudinal integration database for health insurance and labor market studies (LISA), originally from Statistics Sweden (2013), we obtained individual-level socio-economic data such as maternal marital status (married or cohabiting/single), parental education (≤9 years/10–12 years/>12 years), and family disposable income during mother's pregnancy, child's first year of life, and the ninth year of life with adjustment for inflation and family size. Furthermore, a neighborhood deprivation index was used to estimate area-based socio-economic characteristics in the year of birth (Sariaslan et al., Reference Sariaslan, Langstrom, D’Onofrio, Hallqvist, Franck and Lichtenstein2013). Neighborhood was defined by the Small-Area Market Statistics (SAMS) based on regional population density (Statistics Sweden, 2013). Data, including information from Statistics Sweden on welfare beneficiaries, unemployment, immigrants, divorce rate, income, education, residential mobility, and criminal conviction rate were linked with each SAMS unit to calculate a neighborhood deprivation index using principal component analysis. Information on comorbidity with severe chromosome abnormalities, neural tube defects, and other neurological diseases, including epilepsy and cerebral palsy (see Table S1 of the Supplementary Material), was obtained through parent-report in CATSS, as well as from the National Patient Register, according to diagnoses of hospital discharge or outpatient department visits.
Statistical Analysis
Generalized estimating equations (GEE) with exchangeable correlation structure in combination with the Huber–White sandwich estimator for standard errors to adjust for the clustering of observations within twin pairs were used to estimate odds ratios (OR) and 95% confidence intervals (CI) for each neurodevelopmental outcome associated with the 5th to 95th percentile increase in NOx or PM10 on the entire sample (Carlin et al., Reference Carlin, Gurrin, Sterne, Morley and Dwyer2005). We used a directed acyclic graph to determine potential confounders for ORs (Greenland & Brumback, Reference Greenland and Brumback2002). A series of models were run step-wise to assess OR changes by further adjustment for potential confounders; however, only crude and adjusted models, including all potential confounders (p < .20), are presented.
Cut-off values validated in two previous studies were used as outcomes in all analyses (Hansson et al., Reference Hansson, Svanstrom Rojvall, Rastam, Gillberg, Gillberg and Anckarsater2005; Larson et al., Reference Larson, Anckarsater, Gillberg, Stahlberg, Carlstrom, Kadesjo and Gillberg2010). Furthermore, we added a general neurodevelopmental outcome, defined as scoring above any ASD- or ADHD-related cut-off values due to the high co-occurrence of both diseases. Sensitivity analyses were performed using air pollution exposure during each trimester of pregnancy and during child's ninth year of life, controlling for seasonal effect, and by defining cases of comorbidity with severe chromosome abnormality, neural tube defects, and other neurological diseases, including epilepsy and cerebral palsy (see Table S1). Furthermore, a subset of children whose mothers responded at the interview was analyzed to avoid reporting bias among different family members.
The statistic package STATA version 12 (Stata Corp., College Station, TX, USA) was used for all analyses.
Results
Table 1 lists characteristics of the study population. Eligible participants were, on average, aged 10.3 years; 76% of mothers did not smoke during pregnancy, and only 6% of the families had less than 9 years of education. Children with neurodevelopmental disorders were predominantly males, more likely to be born in a lower educated family with at least one parent from Scandinavian countries, exposed to maternal smoking during pregnancy, and with diagnosed comorbidity with severe chromosome abnormalities, neural tube defects, and other neurological diseases, including epilepsy and cerebral palsy. The non-responding twin parents showed some socio-demographic differences compared with those included in the analyses, such as younger maternal age, more single mothers, lower parental education and family income as well as higher neighborhood deprivation.
TABLE 1 Child and Family Characteristics in 9- and 12-Year-Old Twins Born in Stockholm
ASD = Autism Spectrum Disorders; ADHD = Attention Deficit/Hyperactivity Disorders; NPI = Neighborhood Deprivation Index; SD = standard deviation; SEK = Swedish kronor.
aCut-off values of disorders from extended diagnostic criteria: ASD = 4.5 and ADHD = 6.0.
bComorbidity-included co-occurrence with severe chromosome abnormalities, malformations of brain, epilepsy, cerebral palsy, and other neurological disorders. Detailed information on diagnosis codes is listed in Table S1 in supplemental materials.
cThe p-values were presented comparing ASD/ADHD individuals with the ones with neither ASD nor ADHD.
† p < .05; ‡ p < .001.
Figure 2 shows air pollutant levels during pregnancy, and child's first and ninth years of life. Yearly average levels of NOx from local traffic dropped from 12.7 μg/m3 to 5.4 μg/m3 during the observation period, which is reflected in reduced levels from pregnancy/infancy to the ninth year of life. On the other hand, the yearly average levels of PM10 were relatively constant (3.3–4.2 μg/m3). NOx was closely correlated to PM10 (all p-values < .001, r 2 > 0.7) when comparing with the study period as both have local traffic as the major source of air pollution. However, there were only moderate correlations (all p-values < .001, r 2 < 0.4) between pollutants during the ninth year of life and other study periods (see Figures S1a and S1b in the Supplementary Material).
FIGURE 2 Box plot describing the distribution of log-transformed NOx (white) and PM10 (gray) concentrations (μg/m3) from local traffic in study population from mother's pregnancy to child's ninth year of life. Labels of each scale unit on concentration levels (y-axis) were back-transformed. The box and whiskers denoted the 5th, 25th, 50th, 75th, 95th percentile and outlier values of pollutants’ distributions.
The risks of ASD and ADHD using different cut-off values were not consistently associated with exposure to NOx or PM10 at any age (Figure 3 and Table 2). For example, exposure to NOx during the first year of life was not associated with ASD (OR: 0.86, 95% CI: 0.44–1.67) or ADHD (OR: 1.06, 95% CI: 0.71–1.59) after adjusting for child gender, parity, and other relevant covariates. Similarly, exposure to PM10 during the first year of life was not related to ASD (OR: 0.95, 95% CI: 0.56–1.62) or ADHD (OR: 1.06, 95% CI: 0.75–1.52). A lack of association was also observed for exposure to air pollution during pregnancy. Results were similar using the dimensional outcomes for ASD and ADHD (data not shown). It should be noted that there was a substantial overlap between the diagnoses; for example, 82 of the 109 children with ASD also had ADHD.
FIGURE 3 Odds ratios and 95% confidence intervals of neurodevelopmental outcomes by residential address-based NOx (black circles) and PM10 (black hollow diamonds) levels. The spikes represent odds ratios for outcomes of interest and cap lines indicate 95% confidence intervals.
TABLE 2 Crude and Adjusted ORs of Neurodevelopmental Disorders for Twins Born in Stockholm, by Exposure to NOx and PM10 From Mother's Pregnancy to First Year of Life
ASD = Autism Spectrum Disorders; ADHD = Attention Deficit/Hyperactivity Disorders; DSM-IV = Diagnostic and Statistical Manual of Mental Disorders (4th edition); OR = odds ratio; CI = confidence interval.
aEstimates based on crude models.
bModels adjusted for parity, gender, maternal age during pregnancy, maternal smoking during pregnancy, maternal marital status at birth year, parental education, family income, and neighborhood deprivation at birth year.
When exposure to air pollutants for each trimester of the pregnancy controlling for seasonal effect and during the child's ninth year of life was evaluated separately, similar findings were found with no consistent associations for most neurodevelopmental outcomes related to traffic–air pollutant levels (Tables S2–S5 of the Supplementary Material). However, it is noteworthy that an inverse relation was observed between air pollution exposure during the 2nd and 3rd trimesters and ASD, as well as ADHD, using cut-off values based on the DSM-IV criteria. We also did a sensitivity analysis by redefining cases comorbid with chromosome abnormality or neurological diseases (Table S6 of the Supplementary Material). The ORs in those analyses tended to be lower, but still no statistically significant association was found. In sub-analyses, we assessed all twins whose mothers answered the telephone interview from CATSS, and similar findings were found for all outcomes (Table S7 of the Supplementary Material).
Discussion
This study did not indicate an association between exposure to NOx or PM10 from traffic during pregnancy or the 1st year of life and neurodevelopmental disorders in children. For specific subgroups and diagnoses, there were some associations but no consistent patterns were evident. This also holds true for analyses related to exposure during certain time windows.
There is limited evidence on air pollution exposure and neurodevelopmental disorders in children (Becerra et al., Reference Becerra, Wilhelm, Olsen, Cockburn and Ritz2013; Calderon-Garciduenas et al., Reference Calderon-Garciduenas, Engle, Mora-Tiscareno, Styner, Gomez-Garza, Zhu and D’Angiulli2011; Dix-Cooper et al., Reference Dix-Cooper, Eskenazi, Romero, Balmes and Smith2012; Guxens et al., Reference Guxens, Aguilera, Ballester, Estarlich, Fernandez-Somoano and Lertxundi2012; Jung et al., Reference Jung, Lin and Hwang2013; Morales et al., Reference Morales, Julvez, Torrent, de Cid, Guxens, Bustamante and Sunyer2009; Siddique et al., Reference Siddique, Banerjee, Ray and Lahiri2011; Volk et al., Reference Volk, Hertz-Picciotto, Delwiche, Lurmann and McConnell2011; Reference Volk, Lurmann, Penfold, Hertz-Picciotto and McConnell2013; Vrijheid et al., Reference Vrijheid, Martinez, Aguilera, Bustamante, Ballester, Estarlich and Project2012; Windham et al., Reference Windham, Zhang, Gunier, Croen and Grether2006). Windham et al. (Reference Windham, Zhang, Gunier, Croen and Grether2006) reported a positive relation between the distribution of hazardous air pollutants at birth addresses and ASD among children in California. Other studies in California found that living close to freeways and traffic-related air pollution in mother's late pregnancy or child's first year of life was associated with an increased risk for autism (Volk et al., Reference Volk, Hertz-Picciotto, Delwiche, Lurmann and McConnell2011, 2013). Siddique et al. (Reference Siddique, Banerjee, Ray and Lahiri2011) compared children living in the New Delhi (India) urban area with children living in rural areas and showed that ADHD was positively correlated with current PM10 levels. Air pollutants may induce systematic inflammation, which could be a possible mechanism mediating these effects (Block & Calderon-Garciduenas, Reference Block and Calderon-Garciduenas2009; Calderon-Garciduenas et al., Reference Calderon-Garciduenas, Solt, Henriquez-Roldan, Torres-Jardon, Nuse, Herritt and Reed2008).
The results of our study did not indicate that air pollution has an effect on the risk of neurodevelopmental disorders, even when time windows were considered during fetal life and infancy. The apparently discrepant results compared with some earlier studies could have several explanations. First, relatively low levels of air pollution may contribute to the absence of an association and make it difficult to compare with other study settings. For example, the local traffic-related PM10 concentrations during participants’ first year of life in Stockholm was only 3.9 μg/m3 and the long-range transported PM10 in this part of Sweden has a yearly average level of around 10 μg/m3 (Gidhagen et al., Reference Gidhagen, Omstedt, Pershagen, Willers and Bellander2013). The roof top levels for PM10 in central Stockholm have been relatively constant during 1994–2012 (Burman & Norman, Reference Burman and Norman2013). However, these levels are considerably lower than in the study areas of California described above (mean value at 25 ± 7.2 μg/m3 in one study and 36.3 ± 6.1 μg/m3 in another study; Becerra et al., Reference Becerra, Wilhelm, Olsen, Cockburn and Ritz2013; Volk et al., Reference Volk, Lurmann, Penfold, Hertz-Picciotto and McConnell2013). Furthermore, associations may exist between the socio-economic status at individual or neighborhood level and the risk for neurodevelopmental or behavioral problems (Bhasin & Schendel, Reference Bhasin and Schendel2007; Flouri et al., Reference Flouri, Mavroveli and Tzavidis2012). Maternal smoking correlates with socio-economic factors such as education and income (Kabir et al., Reference Kabir, Connolly and Alpert2011; Laaksonen et al., Reference Laaksonen, Rahkonen, Karvonen and Lahelma2005) and may contribute to this association. The earlier studies (Calderon-Garciduenas et al., Reference Calderon-Garciduenas, Engle, Mora-Tiscareno, Styner, Gomez-Garza, Zhu and D’Angiulli2011; Guxens et al., Reference Guxens, Aguilera, Ballester, Estarlich, Fernandez-Somoano and Lertxundi2012; Siddique et al., Reference Siddique, Banerjee, Ray and Lahiri2011; Volk et al., Reference Volk, Hertz-Picciotto, Delwiche, Lurmann and McConnell2011; Vrijheid et al., Reference Vrijheid, Martinez, Aguilera, Bustamante, Ballester, Estarlich and Project2012; Windham et al., Reference Windham, Zhang, Gunier, Croen and Grether2006) did not always adjust for neighborhood deprivation as well as individual socio-economic characteristics and smoking during pregnancy, which suggests that there could be some residual confounding.
We found inconsistent associations between air pollution in late pregnancy and decreased risk of ASD and ADHD using cut-off values based on the DSM-IV criteria. Even though the sample size was relatively large with 3,426 participants, the number of children who scored above the cut-off values for some neurodevelopmental outcomes was low, contributing to the statistical uncertainty of risk estimates.
Strengths of the study include a population-based sample of twins and data linkage to Swedish national registries, which include baseline birth-related and socio-economic information before disease onset. Second, we investigated both ADHD and ASD because of the high degree of comorbidity between the two conditions. Furthermore, we analyzed neurodevelopmental disorders categorically based on the DSM-IV criteria and the additional cut-off values according to previous validation studies (Hansson et al., Reference Hansson, Svanstrom Rojvall, Rastam, Gillberg, Gillberg and Anckarsater2005; Larson et al., Reference Larson, Anckarsater, Gillberg, Stahlberg, Carlstrom, Kadesjo and Gillberg2010). Third, we included different trimesters during pregnancy, first year, and the ninth year of life using the validated dispersion modeling together with data on road traffic emissions, while previous studies reported effects from either pre- or postnatal air pollution exposures.
There are also several potential limitations of the study. One is that the occurrence of neurodevelopmental outcomes may have differed in children participating in CATSS with completed A-TAC assessment and those in the general population. Two Swedish studies found that children of immigrant parents had impaired psychological health (Gillberg et al., Reference Gillberg, Steffenburg, Borjesson and Andersson1987; Magnusson et al., Reference Magnusson, Rai, Goodman, Lundberg, Idring, Svensson and Dalman2012; Van Leeuwen et al., Reference Van Leeuwen, Nilsson and Merlo2012); however, the occurrence of neurodevelopmental disorders in our study was lower in families with both parents from outside of Scandinavian countries. The data linkage to other registers allowed us to acquire additional data on the CATSS non-responders, which indicated that children enrolled in the study had higher familial socio-economic status. Another possible limitation is the assessment of neurodevelopmental outcomes, which might have created some misclassification (Ragland, Reference Ragland1992). Earlier studies mostly attempted to evaluate outcomes as discrete scores; however, our data were highly skewed on all outcomes. Our power was limited for the analyses of sub-dimensional ASD/ADHD measures. Furthermore, for the exposure time measured during child's ninth year of life, the air pollution assessment may actually have occurred after the onset of disease.
Conclusions
We found no support for the hypothesis that traffic-related air pollution is associated with an increased risk for neurodevelopmental disorders in children. Comparatively low air pollution levels and a limited statistical power for some outcomes may contribute to explaining the results.
Acknowledgments
Financial support was provided through the Swedish Research Council for Health, Working Life and Welfare (FORTE 2012-0573), the Swedish Research Council (VR) 2011-3060, VR in partnership with FORTE, FORMAS, and VINNOVA (cross-disciplinary research program concerning children's and young people's mental health), VR through the Swedish Initiative for Research on Microdata in the Social and Medical Sciences (SIMSAM) framework, grant No. 340-2013-5867, HKH Kronprinsessan Lovisas förening för barnasjukvård, and the Strategic Research Program in Epidemiology at Karolinska Institutet.
Supplementary Material
To view the supplementary material for this article, please visit http://dx.doi.org/10.1017/thg.2014.58.
