. The sample for the present analysis was drawn from participants in the Cincinnati Childhood Allergy and Air Pollution Study (CCAAPS). Children enrolled in CCAAPS, a prospective birth cohort, were selected based on their residence at time of birth being either near (< 400 m) or far (> 1,500 m) from a major highway or bus route (LeMasters et al. 2006
). From birth records, newborns were identified from the Cincinnati, Ohio, metropolitan area from 2001 through 2003. Parents who reported allergy symptoms and had a positive skin prick test to one of 15 aeroallergens were eligible to enroll their infant. Children completed annual clinical examinations at ages 1–4 and 7 years. At each of these visits a medical history, physical exam, and skin prick test were obtained. Parents also reported the locations where their child spent at least 8 hr/week during the previous year, including new residences, child care facilities, or relatives’ homes. The CCAAPS study was approved by the University of Cincinnati Institutional Review Board, and parents provided informed consent before their children were enrolled in the study.
Child behavioral assessment
. At the 7-year study visit, parents completed a psychological assessment questionnaire, the Behavioral Assessment System for Children, Parent Rating Scale, 2nd Edition (BASC-2) (Reynolds and Kamphaus 2004
). The BASC-2 consists of 160 questions that are answered by parents as never, sometimes, often, or always. It is designed to assess a child’s adaptive and problematic behaviors in both the community and home settings (Reynolds and Kamphaus 2004
). The BASC-2 was validated for use on a U.S. school-age population. After the parent completed the BASC-2, it was scored using BASC-2 scoring software (BASC-2 PRQ ASSIST™). BASC-2 scores consist of composite and subscale T
-scores with a mean (± SD) of 50 ± 10. All subscale scores were calculated, but Hyperactivity, Attention Problems, Aggression, Conduct Problems, and Atypicality subscales were selected for analysis a priori
because these symptoms often co-exist with ADHD (Cunningham and Jensen 2011
; Volker et al. 2010
). The BASC-2 has three internal validity scores. The F (“Fake bad”) index assesses whether the parent rated the child in an overly negative fashion. A high score (> 6) on this scale suggests that the parent rated the child’s behavior more severely than expected and that the results should be interpreted with caution (Reynolds and Kamphaus 2004
). The Response Pattern (R) index is a count of the number of times an individual item’s response differs from the previous item’s response. A high (> 125) or low (< 66) R index value suggests that parent was not attentive to content of the questions and that the results should be interpreted with caution. The Consistency index identifies situations where the parent provided different answers to questions that are usually answered similarly. A Consistency index score > 17 suggests that the results should be interpreted with caution. The scoring protocol for the BASC-2 allows for no more than two missing items per each rating scale.
-scores were dichotomized into those in the “at risk” range or higher versus those below the “at risk” cut-off T
-score of 59. The “at risk” designation is used to indicate those children who have the potential of developing behavioral problems on a given subscale and require careful monitoring (Reynolds and Kamphaus 2004
Estimation of TRAP exposure
. A complete description of the air sampling at monitoring stations and the land-use regression (LUR) methodology developed and used to estimate traffic exposure for CCAAPS children has previously been described (Ryan et al. 2007
). Briefly, ambient air sampling was conducted on a rotating basis at 27 sampling sites in the greater Cincinnati area from 2001 through 2006, and the average daily concentration of elemental carbon was determined for each. The portion of sampled elemental carbon attributed to traffic (ECAT) was determined by positive matrix factorization and UNMIX receptor models (Hu et al. 2006
; Sahu et al. 2011
). Thus, ECAT serves as a marker of traffic exhaust, primarily diesel exhaust (Sahu et al. 2011
). The final LUR model relating ECAT measured at the sampling sites to land-use and traffic variables had an R2
of 0.73 with elevation, truck traffic within 400 m, and length of bus routes within 100 m as significant independent variables (Ryan et al. 2007
). A time-weighted average daily exposure to ECAT was estimated for each CCAAPS participant during the first year of life based on parental report of locations where the child spent at least 8 hr/week, on average. The average hours per day at each of these addresses was weighted by the proportion of time spent at that location during the first year of life (Ryan et al. 2008
For this study, we used the average daily ECAT exposure over the first year of life as the primary exposure.
. Environmental tobacco smoke (ETS) exposure during pregnancy was estimated by parental report of the number of cigarettes smoked by the mother during each trimester of the pregnancy and averaged to estimate prenatal ETS exposure. Postnatal ETS exposure during the first year of life was also obtained via questionnaire and dichotomized into exposed and nonexposed based on parental report of any cigarettes smoked in the home (Biagini Myers et al. 2012
). The age of the home was used as a surrogate for lead exposure and was dichotomized based on the year the home was built (< 1950 or ≥ 1950). Cotinine measurements were obtained for at least one hair sample collected during a study visit from year 1 through year 4. If more than one measure was available, the earliest obtained sample was used.
At the baseline visit, demographic information was collected. Maternal and paternal education were captured by five categories [did not finish high school, high school or GED (general equivalency diploma), some college or trade school (≤ 3 years), college (≥ 4 years), and graduate school] and dichotomized at completion of high school/GED or less, or some college/trade school and higher. Household income and insurance status (Medicaid) was obtained by caregiver report. Income was dichotomized at $30,000/year, and Medicaid status was dichotomized (yes/no). Race/ethnicity was defined as either African American or non–African American based on parental report. Mothers reported their duration of breastfeeding at the 1-year clinical examination, and it was dichotomized as either > 5 months or ≤ 5 months. Breastfeeding for > 20 weeks was negatively associated with ADHD symptoms and positively associated with executive functioning at age 4 years in a birth cohort of 500 children (Julvez et al. 2007a
). Child care attendance (yes/no) during the first year of life was determined by parental report of their child spending time at a babysitter, child care, or a relative’s home.
. We estimated the association between ECAT exposure and behavior problems using linear regression for continuous BASC-2 subscale scores and logistic regression for dichotomized BASC-2 subscale scores defining children “at risk.” Because of the skewed distribution of ECAT exposure and the relationship between the highest tertile of ECAT exposure and recurrent night cough previously reported in this cohort (Sucharew et al. 2010
), ECAT exposure was dichotomized at the highest tertile versus the lower two tertiles. An exploratory analysis was also performed with ECAT and BASC-2 scores as continuous variables.
Potential model covariates were identified based on previous work within the CCAAPS cohort linking environmental exposures with outcomes, and on established risk factors for childhood behavioral disorders, including sex, race/ethnicity, maternal and paternal education, household income, Medicaid status, duration of breastfeeding, pre- and postnatal ETS exposure, hair cotinine, age of home, and child care attendance (Biederman and Faraone 2005
; Braun et al. 2006
; Julvez et al. 2007a
). Factors that predicted BASC-2 Hyperactivity subscale score in the “at risk” range with p
< 0.1 based on chi-square tests and bivariate logistic regression were entered into multivariable models for each outcome. Colinearity among the potential covariates was also assessed using chi-square, Pearson correlation, analysis of variance, or Kruskal–Wallis, as appropriate. To avoid colinearity among several covariates that could serve as markers for socioeconomic status, only maternal education was chosen to represent socioeconomic status in the multivariable models. To evaluate the potential differential impact of this variable on hyperactivity scores, additional analyses were performed. An interaction term (maternal education × ECAT) was entered into the model and further explored by an analysis stratified by maternal education.
Statistical analysis was carried out using Epi Info 3.5.1 (CDC, Atlanta, GA) and JMP 9 (SAS Institute Inc., Cary, NC).