Asthma is a complex disease that is influenced by environmental factors and is associated with substantial morbidity in childhood.¹ Previous researchers have established a connection between environmental risk factors and poorer respiratory health Several studies have elucidated the harmful effects of pollutants and chemicals such as particulate matter, ozone, nitrogen dioxide^2,3 and agricultural pesticides.^4,5 However, there is limited information on pesticide exposure in non-occupational settings and its effect on pulmonary health in children.

Pesticides are commonly used in the United States in residential settings.⁶ Children are particularly susceptible to pesticides due to more sensitive organ systems and lower capacity to metabolize and eliminate chemicals compared to adults.⁷ Pesticide exposure occurs through multiple routes (dermal, oral, inhalation). The oral route via food intake is considered the most important exposure pathway for the general population.⁸ However, non-dietary and inhalation routes may also be important exposure paths especially for children.⁹ The health risks of pesticides are of particular concern for children with asthma who live in low-income, multi-family housing. These children are more susceptible due to their impaired respiratory health, and more vulnerable because of persistent pest infestation often found in urban, multi-family environments. ^10,11

Although the toxic potential of low-level exposure to pesticides is unclear, pesticide toxicity at high exposures is well known. ^12–14 There is some evidence that organophosphorous and pyrethroid insecticides, and certain herbicides are toxic substances that have the potential to cause long lasting health effects even in low doses. ^15–17 However, research assessing low-level chronic exposure in children, particularly for those at greatest risk, is lacking.

Another research gap is the lack of collective consideration of multiple classes of pesticides; to date, most researchers have focused on a single metabolite or pesticide class ^4,6,9,18 even though pesticide exposures do not occur in isolation. In a study evaluating general population exposure to a wide range of pesticides, Li et al.¹⁹ found that organophosphorous biomarkers accounted for 62–77% of the total daily urinary pesticide concentrations and pyrethroid metabolites accounted for 17–33% of the total daily pesticide concentrations. Other research has shown the general population is exposed to multiple classes of pesticides^8,20,21 and an association with non-respiratory outcomes. ^22–24Pesticides comprise many substances with dissimilar structures and diverse toxicity that may act through different mechanisms. Therefore, it is plausible that not just one, but several mechanisms are involved in the pathophysiological routes between pesticide exposure and respiratory health. Despite this, there are limited data on exposure to mixtures of pesticides on respiratory health in the general popualtion. A recent publication from the Agency for Toxic Substances and Disease Registry concluded that greater-than-additive action on neurologic end points is possible between certain pyrethroids and organophosphorous insecticides, however the relevance of these findings to relatively low environmental exposures to pyrethroid and organophosphorous insecticides is not well understood.²⁵ Therefore, although not conclusive, there is biological plausibility for the combined negative effects of pesticide exposure at least on neurologic endpoints. The evidence for respiratory outcomes has shown 2,4-dichlorophenoxyacetic acid (2,4-D), and pyrethroids and organophosphorous metabolites to be associated independently with respiratory outcomes.^26–30 However, there is limited evidence on their combined effects, especially in the general population. Research assessing potential health effects of pesticides would benefit from considering both the individual and mixture effects of common pesticides on asthma. The current study fills this important data gap.

The objectives of this study were to quantify urinary biomarker concentrations of select commonly used pesticides in children ages 6 to 13 years who reside in an urban environment and to assess the association between pesticide exposure biomarkers and various measures of asthma morbidity. The hypothesis is that greater concentrations of urinary pesticide biomarkers (individual and mixtures) are associated with more asthma attacks, greater likelihood of asthma related healthcare utilization, and more airway inflammation.

Characteristics of the 162 children at baseline are described in Table 1. Forty-two percent of children were from New Orleans, 28% from Cincinnati, and 30% from Boston. The median age was 9.5 years, a majority identified as non-Hispanic black (71%), and most lived in households with an annual income of less than $25,000 per year (91%). All children had medical insurance. At the baseline visit, 30% of children reported having had an asthma attack in the previous three months, 11% reported asthma related health care utilization in the previous three months, and 16% had evidence of eosinophilic inflammation (FeNO >35 ppb).

Detection frequencies of pesticide biomarkers varied over the 12-month follow-up period (Table 2). The most commonly detected pesticide biomarkers were PNP (100%) followed by TCPy (99%). Only one pyrethroid biomarker was commonly detected (3-PBA) with a detection frequency of 93.2%. The herbicide 2,4-D was detected in 71% of urine samples. IMPY, 4F-3PBA, and trans-DCCA were detected in <60% of samples and will not be discussed further.

Geometric mean concentrations of the pesticide biomarkers are presented in Figure 1. TCPy had the highest concentration at each time point (Baseline: 1.33 μg/g creatinine, Month 6: 1.23 μg/g creatinine, Month 12: 1.44 μg/g creatinine) and 2,4-D had the lowest concentration (Baseline: 0.33 μg/g creatinine, Month 6: 0.31 μg/g creatinine, Month 12: 0.24 μg/g creatinine). The ICCs over the year ranged from 0.06 for TCPy to 0.43 for 3-PBA. The Spearman correlation between pesticide biomarkers concentrations at each time point are presented in Figure 2. Correlations ranged from negative 0.14 between PNP at baseline and 3-PBA at month 12 to 0.42 between 3-PBA at baseline and 3-PBA at month 6. Convenience and FMV spot sample concentrations of PNP, TCPy, 3-PBA and 2,4-D were significantly correlated at each time point. Correlations ranged from 0.39 (PNP at baseline) to 0.75 (TCPy at baseline).

Pesticide median concentrations differed significantly by study site (Table S1). Children in Boston had the highest median concentration of PNP, whereas children in Cincinnati had the highest median concentrations of TCPy and 3-PBA. New Orleans children had the highest median concentration of 2,4-D. Comparing children in the GHS to those from NHANES, a nationally representative sample, GHS children had significantly lower pesticides biomarker concentrations (Table S1).

Results of bivariate analyses can be found in Table S2. Daily medication use, and dust mite (Dermatophagoides pteronyssinus and D. farinae) and cockroach (Blatella germanica) sensitization were associated (p<0.20) with the number of asthma attacks in the previous three months. BMI, daily medication use, dust mite sensitization, season, and age were associated with health care utilization. Sex, race, household income, sensitization to dog, dust mite (both D. pteronyssinus and D. farinae), cockroach, cat, and mouse, site, season and age were associated with FeNO.

Unadjusted and adjusted associations for each pesticide biomarker are presented in Table 3. For asthma attacks, associations with urinary biomarkers were not significant in unadjusted or adjusted models. In unadjusted models for asthma related HCU, TCPy and 3-PBA were inversely associated with HCU while PNP and 2,4-D were positively associated. None of the associations reached statistical significance. In models adjusting for BMI, medication use, dust mite sensitization, season and age, PNP and 3-PBA were positively associated with health care utilization (PNP: OR 2.05 95% CI: 0.76–5.52; 3-PBA: OR: 1.07 95% CI 0.88–1.29) but did not reach statistical significance. In models evaluating the association between pesticide biomarkers and FeNO, PNP and 2,4-D were significantly associated with decreased FeNO (PNP: −17.4%; 2,4-D:−19.74%).

Using JWQS regression we evaluated exposure to a mixture of the four most commonly detected pesticides biomarkers, TCPy, 2,4-D, 3-PBA and PNP,on asthma attacks and asthma related health care utilization over the year. The mean number of asthma attacks was 1.5 (range 0 – 20) and 20% of children reported at least one incidence of asthma related health care utilization during the year. The mixture index was not associated with asthma attacks in unadjusted models (IRR: 0.95 97.5% CI: 0.78,1.16) or in models adjusting for daily medication use, and sensitization to dust mites and cockroach (IRR: 0.98 97.5% CI: 0.83,1.17). In unadjusted models the pesticide mixture index was associated with HCU (OR: 1.56 97.5% CI: 1.08,2.27) with the greatest contribution coming from 3-PBA (55%), the mixture index remained positively associated with HCU in models adjusting for BMI, daily medication use, dust mite sensitization, season at baseline and child’s age (OR: 1.40 97.5% CI 0.86,2.29) although not statistically significant. 3-PBA continued to contribute the greatest weight to the index (45%) followed by 2,4-D (28%), then PNP (15%) with TCPy contributing the least (10%) (Table 4.).

We found that children’s pesticide biomarker concentrations vary by geographic location, type of pesticide, and collection time point. The ICC revealed poor reproducibility between concentrations of urine pesticide biomarkers signifying that over the year of follow-up, pesticide biomarker concentrations varied considerably. A notable finding is that pesticide biomarker concentrations in this study population at each site were lower than in children from the general U.S. population (NHANES 2013–2014). We did not find significant associations between any of the individual pesticide biomarkers and the respiratory outcomes in the expected direction. We found a protective effect for PNP and 2,4-D on FeNO, which may relate in part to the fact that PNP is a biomarker of other compounds in addition to pesticides, this association may also simply be a spurious association in our data set that would need confirmation in future studies. The variability across time and geography may help explain inconsistent associations between pesticides and respiratory health currently found in the literature although we cannot rule out other factors for these associations.

Examining the effect of individual pesticide biomarkers on asthma morbidity, our results suggest higher concentrations of PNP and 3-PBA result in a greater likelihood of asthma-related health care utilization, although neither association reached statistical significance. Of the individual metabolites assessed, the strongest associations were seen with PNP. Significant negative associations were found between PNP, 2,4-D and FeNO. Associations with asthma attack were small and non-significant.

Our results add to the research evaluating pesticides and asthma. In a recent systematic review examining the respiratory and allergic effects in children exposed to pesticides,⁴⁰ of the seventeen studies evaluating organophosphorous pesticides or multiple pesticide exposure, eleven reported positive associations, four found no associations, and two had no clear findings. An important note is that a majority of the studies were conducted in populations exposed via agricultural proximity or work. Of the three studies that evaluated pesticides in populations not including agricultural areas, only one included measurement of a biomarker of exposure.⁴¹ In that study the authors found no clear association between asthma and biomarkers of diakylphosphates (DAP) a non-specific marker of OPs. The remaining two studies evaluated pesticide exposure via questionnaire and found children exposed to pesticides had higher odds of having respiratory disease (OR= 1.71; 95% CI: 1.20–2.43)⁴²; and asthma diagnosis before 5 years was significantly associated with exposure to pesticides in the first year of life (OR: 2.39; 95% CI: 1.17–4.89)⁴³. So, it appears that exposure to pesticides is related to impaired respiratory health however, differences in which pesticides were evaluated, methods of exposure assessment, and outcomes make direct comparisons across studies difficult.

We found a significant positive association between the pesticide mixture index and asthma-related health care utilization in unadjusted models. In adjusted models, the association remained positive, however the effect was attenuated and crossed the null. The pesticide index was predominately weighted by the pyrethroid biomarker 3-PBA and the herbicide 2,4-D which may indicate these biomarkers as the drivers of the effect. Therefore, evaluating the association of not only the individual metabolites but the combined exposures, which most closely mimics real world exposure, is important. Although not significant the results suggest there may be an increased risk of asthma related health care utilization even among children with biomarker concentrations lower than those in the general population. Therefore, there is a need to increase our understanding of the potential effects of pesticides in less-studied populations, such as those residing in low-income, multi-family, housing.^44,45

The exact mechanism through which pesticides may affect asthma morbidity is unknown, although pesticide exposure may contribute to the exacerbation of asthma by irritation, inflammation immunosuppression, or endocrine disruption.⁶ The different mechanisms, classes, and mixtures of pesticides have made it difficult to assess individual effects of pesticide exposure on asthma morbidity. This may explain the varying results observed in the current study and previous research when looking at different pesticides and respiratory outcomes.

An unexpected finding is the negative association between FeNO and PNP and 2,4-D. One possible explanation for PNP may be related to exposure of cigarette smoke in participants’ homes. PNP is also a metabolite of nitrobenzene which is a compound found in tobacco smoke.⁴⁶ Urinary cotinine measurements suggest there was smoking in the homes of some of the participants and we know that smoking may induce a decrease of exhaled NO levels in both healthy persons and those with asthma. Several mechanisms, such as an inhibited bronchial epithelial iNOS-related production of NO or an increased catabolism of NO, have been postulated to explain this decrease.⁴⁷ Based on the presence of nitrobenzene (a parent compound of PNP) in tobacco smoke and the biologic relationship between tobacco smoke and FeNO, it is plausible that in our study PNP was acting as a surrogate for tobacco smoke and therefore decreased FeNO levels. However, this study was not designed to evaluate sources of pesticides.

The study has a few limitations. Although urinary biomarkers of pesticide exposure are considered the gold-standard for exposure assessment, the relatively short biological half-life of these biomarkers may lead to misclassification of exposure over longer time periods. To overcome this limitation, we used biomarker concentrations from six spot samples to represent pesticide exposure over one year. This is important because we found that pesticide biomarker concentration is not consistent throughout the year. Therefore, the repeated measures of biomarkers strengthen the results of the study. The study was also limited by a relatively small sample size and low variation in study outcomes which may have resulted in non-significant findings. In addition, the use of simple imputation for pesticide biomarker concentrations below the LOD may result in impaired estimates of variances and covariances. Biomarker concentrations were adjusted for creatinine although there is currently controversy over the most appropriate method to adjust for creatinine;⁴⁸ the decision to adjust biomarkers concentrations for creatinine was made to ensure the ability to use all 6 urine samples over the course of the year while still considering the potential urinary dilution. A sensitivity analysis was conducted with unadjusted concentrations and differences in effect estimates were not significant (data not shown). Finally, the study sample may not be generalizable to all children.

The current study has several strengths. Many epidemiological studies of pesticide exposure have used questionnaire-based measurements, job titles, or relative distance to pesticide sources as a surrogate of pesticide exposures for individuals.⁴ Of the studies conducted in children, many used only one or two bio-specimens to assess exposure. The epidemiological studies supported by quantification of biomarkers in urine, like the current study, lead to more (highly) validated conclusions on the associations between human exposure to organophosphorous or pyrethroid insecticides and health outcomes compared to those using questionnaire-based measurements. ¹⁸ In addition, the study evaluated not only the individual metabolites but their mixture.

Our hypothesis was not entirely supported. In this prospective study, there is suggestive evidence that urinary concentrations of pesticide biomarkers are positively associated with asthma related health care utilization. However, we saw no effect on asthma attacks and protective effects with a biomarker of airway inflammation. The study provides unique insight into the potential adverse association between exposure to pesticides and asthma morbidity in children thanks to the repeated measures design, biomarker-based exposure assessment, and assessment of pesticides exposure as a mixture. Future studies with larger sample sizes, more frequent longitudinal assessment of pesticide biomarkers, and different combinations of pesticides could better explore the relationship between pesticide exposure and asthma morbidity.