For this analysis we hypothesized an association between PCBs and attention-associated behaviors in infancy. Previous literature suggests that prenatal PCB exposures affect attention later in childhood (Grandjean et al. 2001
; Jacobson and Jacobson 1996
; Peper et al. 2005
; Vreugdenhil et al. 2004
). A study of children born to Lake Michigan fish consumers reported associations between pre-natal PCB exposure and poorer performance on a Digit Cancellation task, which indicates difficulty with focused attention and concentration, among 11-year-old children who had not been breast-fed (Jacobson and Jacobson 2003
). Associations were also reported among these same children between PCBs and poorer freedom from distractibility, a subscale of the Wechsler Intelligence Scales for Children (Jacobson and Jacobson 1996
). Associations were not detected, however, between PCBs and measures of sustained attention among children in this study at 11 years of age, or at an earlier assessment made at 4 years of age (Jacobson and Jacobson 2003
; Jacobson et al. 1992
). The Faroe Islands study found associations between PCBs and attention measured by a continuous performance test among children 7 years of age only in the context of high mercury exposure, suggesting a potential interaction between these contaminants (Grandjean et al. 2001
). A Dutch study found associations between PCBs and sustained attention among 9-year-olds, measured by a continuous performance tests (Vreugdenhil et al. 2004
). Another study among adults exposed to PCBs from a contaminated building found associations with attention as well as distractibility (Peper et al. 2005
The present analysis was designed to investigate whether an association between organochlorines and attention could be detected in early infancy. We therefore focused on a priori
selected individual NBAS items, including supplementary items (Brazelton and Nugent 1995
), as well as cluster measure outcomes that we believe reflect attention-associated skills. In addition, there is substantial overlap between childhood attention disorders [for example, the diagnosis of attention deficit hyperactivity disorder (ADHD)] and the diagnosis of motor impairment as occurs, for example, in developmental coordination disorder (Gillberg 1998
; Kadesjo and Gillberg 1998
). Independent of co-morbidities, children with ADHD may also have poor movement ability (Gillberg 2003
; Pitcher et al. 2003
), so NBAS measures of motor function were included in our analyses.
Our results show consistent inverse associations between PCB, TEQ, and DDE cord serum levels and attention-related NBAS outcomes, including alertness, quality of alert responsiveness, and cost of attention (). The NBAS supplementary items analyzed (quality of alert responsiveness, cost of attention) were designed to capture qualitative aspects of the infant’s ability to attend to visual and auditory stimuli. For example, quality of alert responsiveness is designed to measure “processing” alertness, not just simple awake, eyes-open states. Given this specified purpose, it is notable that the association with cord serum PCBs and TEQ was particularly strong for supplementary items (). These supplementary items were not included in the NBAS manual until 1995, well after most previous studies of PCBs and NBAS had been conducted, thereby affording an opportunity to assess a more inclusive set of attention measures in this analysis than had been possible previously.
Inverse associations were less pronounced for state- and motor-associated outcomes ( and ). We found little evidence for an association with the three cluster outcomes analyzed, which did not corroborate results of the previous studies conducted in Oswego, New York, and in North Carolina that found associations between PCBs and the habituation cluster (Rogan et al. 1986
; Stewart et al. 2000
The positive association observed between organochlorines and spontaneous activity is consistent with hyperactivity–impulsivity observed in experimental animal models of early-life PCB exposure (Berger et al. 2001
; Holene et al. 1998
; Rice 2000
). Impulsive responding has been reported among humans exposed to PCBs. A study designed to dissociate response inhibition from attention found associations between PCBs and impulsive responding among children 8 and 9.5 years of age but not with sustained attention (Stewart et al. 2005
). These results are consistent with earlier studies of this cohort in which an association between PCBs and response inhibition was reported at 4.5 years of age (Stewart et al. 2003
). The Michigan study also found associations between PCBs and impulsivity among 11-year-olds who had not been breast-fed (Jacobson and Jacobson 2003
We examined whether organochlorines impaired the infant’s ability to recover from the birth experience by assessing changes in performance between the birth exam and the 2-week exam. Recovery is informative of an infant’s ability to cope and adapt to the extrauterine environment, and was an originally intended application of the NBAS (Brazelton and Nugent 1995
). Our findings for recovery were limited in power, however, because data were missing for either exam (both were required).
PCBs are a heterogeneous mixture of congeners, potentially with different toxicologic modes of action. A considerable strength of this study was our ability to measure up to 51 different PCB congeners. Though not presented in the results, effect sizes for the sum of 51 congeners were similar, though slightly attenuated, compared with the sum of 4 PCBs; these 4 congeners are more prevalent and likely measured with more accuracy than the sum of 51 congeners, which may explain the more attenuated effects found for the sum of 51 congeners. The correlation between the sum of 51 and the sum of 4 congeners was also quite high (0.91), and we therefore presented only the sum of the 4 congeners, which we believe adequately represents the sum of all congeners.
To take advantage of congener-specific information, we grouped congeners into structurally related classes—dioxin-like PCBs (TEQ) and non-dioxin-like PCBs (sum of 4 congeners)—that represented potentially different biologic mechanisms in their effect on neurodevelopment. That we did not find a difference in the effect of these two groups on NBAS outcomes was not surprising, because these groups were highly correlated in our study. Categorizing PCBs into such congener classes is important, however, for understanding the mode of action of these different compounds and their relative neurotoxic potency, both of which have implications for risk assessment (Schantz et al. 2003
Serum PCB levels in the New Bedford study population were generally lower than in other birth cohort studies of PCB exposure and likely reflect more general population levels. A comparison of 10 studies of diverse locations and birth years found that the New Bedford cohort was at the low range of PCB levels (represented by PCB congener 153) relative to other populations (Longnecker et al. 2003
). This would have reduced power to detect associations; that we still found consistent associations with attention indicates potential for effects at low PCB levels found in the general population.
It is unclear how much can be learned of an individual’s long-term behavior from observations in infancy. Previous studies find that infant attention is predictive of intelligence in childhood and adolescence (Sigman et al. 1986
). A study of temperament in infancy (5–6 months) and toddlerhood reported significant prediction of impulsivity and inattention among 8-year-old children (Olson et al. 2002
). Other studies did not find infant temperament alone to be predictive, but rather that the combination of infant temperament with parental attitude or perception was predictive of behavioral problems later in childhood (Cameron 1978
; Oberklaid et al. 1993
; Wasserman et al. 1990
). These studies suggest that infant temperament and the interaction between the infant and his or her environment are important determinants of later behavioral outcomes.
How well the NBAS predicts attention later in childhood is also uncertain. Some studies report poor correlation between NBAS and behavioral outcomes later in childhood (Risholm-Mothander 1989
; Sameroff et al. 1978
), whereas another study found behavioral and reflex clusters to be good predictors of later developmental disabilities among a high-risk population (Ohgi et al. 2003
). An important area for future study will be to examine whether these PCB-associated decrements in attention-related skills in infancy are transient or predict attention-related problems later in childhood.
We did not assess maternal IQ at birth, though we did measure maternal IQ with the Kaufman Brief Intelligence Test (Kaufman and Kaufman 1990
) on a subset of the children that were available for testing at 8 years old. When we included maternal IQ as a covariate the exposure–outcome effect was slightly stronger, although the precision was reduced because of the smaller number of children on whom maternal IQ was available. In addition, maternal IQ was not predictive of any of the analyzed NBAS items. Final estimates are therefore reported without adjustment for maternal IQ.
This analysis was limited to subjects who were administered an NBAS exam (539 had an exam around the time of birth, 542 had an exam 1–3 weeks after birth, and 408 had both exams). This process resulted in reduced study power, particularly for the recovery analysis, which required both exams. However, we do not expect that it resulted in bias as organo-chlorine levels and covariates for these groups were very similar (data not shown).
Every effort was made to schedule infant exams in the second week of life; however, it was difficult to examine young infants at such a narrow age range in a nonclinical, population-based study. The use of a broad time window for the 2-week exam may have introduced some random error into our estimates. Exposure effect estimates for NBAS exams conducted for a more narrow window (10–17 days) were very close to estimates from the broader time window (5–22 days), suggesting that including a broader age range did not bias our results.
We evaluated a number of outcomes for this analysis as well as several categories of PCB congeners and DDE. Multiple comparisons were therefore performed and may be a limitation when interpreting positive findings. Our observation of internal consistency, however, particularly for the attention-associated outcomes, suggests that our findings were probably not attributed to chance.
In summary, we found evidence for an association between low-level prenatal organo-chlorine exposure measures and attention in early infancy. This observation is particularly notable given both the low-level PCB exposure in our study population and the limitations of behavioral assessments in young infants. The longitudinal design of the New Bedford cohort will enable us to further determine whether a) poor attention-associated skills in early infancy (as measured by the NBAS) persist in later childhood—that is, whether a behavioral pattern seen in infancy predicts later childhood behavioral skills; and b) the observed association of prenatal organochlorine exposure with attention in infancy persists in later childhood. Identifying attention-related deficits as early as infancy, and identifying potentially remediable risk factors for such deficits (for example, PCB exposure), allows for early intervention (and ultimately prevention efforts), which may be important for promoting healthy subsequent neurodevelopment.