Higher pyrethroid exposure was associated with an elevated odds of ADHD and ADHD symptoms in this nationally representative sample of US children. Our results are consistent with the findings in animal models [21–24], but contrast with a study by Quiros-Alcala et al. which also used an NHANES sample but did not find pyrethroid exposure to be significantly associated with ADHD [26]. This contrasting result may be due to differences in the primary outcome definition. The Quiros-Alcala et al. study’s primary outcomes were based on parental report of ADHD and/or parental report of a learning disability; they did not incorporate diagnosis of ADHD or ADHD symptom counts as determined by the DISC, which may represent more sensitive measures. Additional differences between the Quiros-Alcala et al. study and the present study include the age range (6–15 vs. 8–15 year old) and NHANES study years (1999–2002 vs. 2001–2002). Another prior study in children found an association between postnatal pyrethroid exposure and a general parent-report measure of behavior problems, but did not find a specific association with hyperactivity/inattention [25], perhaps due to the diminished sensitivity of their 5-item ADHD symptom measure (compared to our DSM-IV-based measure of 18 symptoms).

Our results suggest that pyrethroid pesticide exposure may be more strongly associated with hyperactive-impulsive than inattentive symptoms. This is consistent with previous animal studies demonstrating that pyrethroid exposure during gestation and early childhood results in hyperactivity and impulsive-like behavior [19, 23]. Should the link between pyrethroid exposure and hyperactivity-impulsivity in children be replicated, these findings might represent a potential pyrethroid-specific ADHD phenotype. Efforts to characterize etiologic-specific ADHD phenotypes are paramount not only to elucidating the biologic basis of ADHD: they may ultimately prove useful in defining subgroups of children with ADHD with a distinct course, comorbidity profile, and intervention response, and thus form the basis for targeted clinical and treatment efforts [4].

We observed sex-specific effects of pyrethroid exposure. Urinary pyrethroid biomarkers were associated with increased odds of an ADHD diagnosis and number of ADHD symptoms for boys but not girls in logistic and poisson regression as well as spline analyses. The sex*pyrethroid interaction had a borderline significant p-value of 0.09 when 3-PBA was modeled as a logarithmically-transformed continuous variable. In contrast, the interaction was not statistically significant (p = 0.68) when 3-PBA was modeled as a dichotomous variable [above versus below the limit of detection]). Of note, we had reduced statistical power to detect sex*pyrethroid interactions in dichotomous models due to relatively small numbers of participants having the ADHD diagnosis. This pattern suggesting possible sex-specific effects is consistent with prior animal studies that reported greater impulsive and hyperactive symptoms in exposed male but not female mice [23, 24]. Further, these animal studies suggest that male-specific effects of pyrethroid exposure on the brain dopamine system underlie the observed sexual dimorphism in behavioral effects [23, 24]. It should be noted that Oulhote et al. found a stronger adverse association between pyrethroid exposure and a measure of total behavioral difficulties as well as oppositional/conduct-specific problems in girls compared to boys [25], although the contrast between their findings and our own is not direct given that the sex-specific effects were observed on different outcomes. Quiros-Alcala et al. did not find any significant interactions (all p > 0.10) between sex and pyrethroid exposure, and it is not clear if they observed any patterns suggesting possible sex-specific effects despite the failure to meet thresholds for statistical significance. Certainly, further study of sex-specific vulnerability to pyrethroid exposure is warranted, given mounting evidence that the adverse effects of some ADHD risk factors may be accentuated or limited to certain groups [40].

Strengths of this study included our use of a nationally representative sample of the U.S. population, such that pyrethroid exposure levels are more likely to represent the experience of a broad range of American children and adolescents compared to regionally based studies. In addition, our use of both linear and spline models allowed us to explore more carefully the dose–response relationship between childhood pyrethroid metabolite levels and ADHD. Our exploration of the relationship between pyrethroid levels and ADHD symptom counts – in addition to ADHD diagnostic status -- also enable a fuller appreciation of the unique ADHD sub-domains that might be more affected by pyrethroid exposure. However, it should be noted that our Poisson models show some evidence of over-dispersion and therefore significance may be overestimated.

An additional limitation of our study is that pyrethroid exposure was assessed using 3-PBA concentrations in a single spot urine sample. Given that pyrethroids are non-persistent and rapidly metabolized, serial measurements would provide a more accurate measurement of typical exposure [41], and are recommended in future studies. It should also be noted that 3-PBA detected in child urine may not be due entirely to exposure to parent pyrethroid compounds, as there may also be a minor contribution from exposure to a byproduct formed after pyrethroids have been hydrolyzed in the environment. Thus, we expect that some degree of exposure misclassification occurred in this study; if it is non-differential in nature, the tendency would be to bias toward the null, making results an under-, rather than overestimate of the true association [42, 43]. Subsequent studies should also examine the association between measures of prenatal pyrethroid exposure (which are not available in NHANES) and later development of ADHD, as data from animal studies suggest that both the prenatal and postnatal periods may represent susceptible phases for pyrethroid neurotoxicity [18–23]. Another caveat to our study is that we categorized urinary pyrethroid concentrations using the LOD as a cut point because the LOD represents an easily interpretable threshold. However, to guard against the assumption of constant effect sizes without our categories, we also conducted analyses using a restricted cubic polynomial spline, which makes no assumptions about the shape of the dose–response relationship (see Figs. 1, 2, 3).

Misclassification of ADHD may also have played a role in analyses for which case ascertainment included caregiver report of the child’s prior diagnosis of ADHD, as we cannot verify that doctors in community-based settings used DSM-IV criteria when establishing ADHD diagnoses [44]. However, our results were similar in direction and magnitude when ADHD diagnosis was based on the standardized, valid, and reliable DISC instrument (see Additional file 1: Table S1), although the findings were less precise (likely due to a reduction in cell size for this outcome), and the DISC is a caregiver-completed measure that does not incorporate teacher reports of ADHD symptoms. Furthermore, we caution that correlation does not prove causation in this cross-sectional study. Although we adjusted for a number of potential confounders, we were unable to control for genetic and other factors (e.g., diet, other aspects of the caregiving environment) that may account for the observed associations and could in fact represent reverse causality. Nonetheless, prospective animal studies in which case and control subjects had identical genetic lineages and rearing environments, but differed only in the pesticide exposure, document pyrethroid effects on hyperactivity and impulsivity [18–23] that corroborate our findings.