Birth certificate data were obtained from the Bureau of Vital Statistics, Arizona Department of Health Services and collected from January 1, 2006, to December 31, 2013. After restricting the population of 733,571 births to singleton, live-birth deliveries from individuals aged 15 to 50 years old, 707,314 birth records were retained. We excluded 14,591 births with missing occupational data, and 71,438 missing covariate data, resulting in data from 623,185 deliveries (Figure 1). This study was approved by the Human Subject Protection Program of the University of Arizona.

We utilized birth certificates from the period 2006 to 2013, since those years included occupational data for both mothers and fathers. Occupational data were open field, self-reported parental occupations in either English or Spanish, with agricultural workers classified using information on industry and/or occupation. Agricultural workers included: farmers, farmworkers (trabajador agrícola), field laborers (obrero del campo), field equipment operators, foremen/supervisors, growers, green house/nursery workers, harvesters, horticulturists, irrigation workers, cultivators (cultivador), packers (empacadora), sorters, graders, as well as pesticide handlers (mixers, loaders, and sprayers). As pesticide exposures could vary significantly within farm employment categories, we omitted from the definition occupations as florists, gardening/landscaping, office administration, pest control/termite/exterminators, pesticide manufacturers, produce inspectors, livestock, or dairy workers. Agricultural occupation exclusions were adopted from the U.S. Bureau of Labor Statistics, group 45–2000.²⁴ Agricultural households were classified as having either the mother, father, or both reporting an occupation as agriculture or farming. Births missing both parental occupations were excluded from this analysis (n=14,983).

A newborn weighing less than 2,500 grams (g) was defined as low birth weight (LBW). For macrosomia, we used a graded scale to reflect increasing clinical morbidity at higher values. Grade 1 macrosomia was classified as more than 4,000g, grade 2 at 4,500g, and grade 3 at 5,000g. Large for gestational age (LGA) was defined as birth weight greater than the U.S. sex-specific 90th, 95th, and 97th percentile of weight for each week of gestation. Postterm birth (PTB) was defined as a delivery at 41 weeks and beyond, and a delivery occurring at ≤37 weeks was classified as a preterm birth (PTB). Gestational age estimate (weeks) and birth weight (grams) were numeric fields (and not checkboxes). APGAR is a standard assessment used to evaluate the physical fitness of neonates after 5-minutes of being born, and a score ≤7 was defined as low.

Covariates were selected a priori. The final models were adjusted for the following covariates: maternal age at delivery (years), maternal education (≤high school diploma, high school diploma/GED, college, and professional degree), race/ethnicity (White, Latina/Hispanic, Black, Other), previous number of living children, and having any diabetes (yes/no).

Baseline characteristics were presented as means (standard deviation [SD]) for continuous variables and numbers (percentage) for categorical variables. Multivariate logistic regression was used to calculate odds ratios (OR) and 95% confidence intervals (95% CI) to estimate associations of agricultural household exposure status with adverse neonatal outcomes. Exposure status was determined by the number and type of parents in farm work (mother, father, both). Adjusted ORs (aOR), which separately incorporated information on mother and father employment status, were also calculated.

We conducted several sensitivity analyses. First, since smoking and gestational weight gain were not included in the primary analyses due to the high levels of missingness in birth records (>10%), we evaluated smoking and gestational weight gain as potential confounders in the subset of births with these data. Second, due to concern about the potential impact of pre-existing chronic hypertension, we excluded these individuals from the model. Third, we re-defined agricultural workers to include primarily field workers and farmworkers, to exclude workers who were primarily administrative and/or with less physical tasks (e.g., horticulturists, supervisors/foremen, and growers). For these analyses, models were evaluated with reduced data subsets. A fourth sensitivity analysis examined the potential effect modification by fetal sex or maternal age (≤ 21 years old versus ≥ 35 years old). Interaction terms were added to the final models and the criterion of alpha of p < 0.05 for evidence of interaction between the variables using a likelihood ratio test was applied. Statistical analyses were conducted using RStudio version 1.1.453.

Table 3 shows estimates of the magnitude of associations between agriculture worker household status and various neonatal outcomes. Newborns born to a household where at least one parent worked in agriculture significantly increased the odds of grade 1 macrosomia (aOR 1.15, 95% CI: 1.05, 1.26), and grade 2 macrosomia (aOR 1.38, 95% CI: 1.11, 1.71). While the effect estimate continued to increase for grade 3 macrosomia (aOR 1.48, 95% CI: 0.79, 2.77), this association was no longer statistically significant, likely due to the small numbers of births from agricultural households (n=10). Similarly, parental employment in agriculture increased the likelihood of LGA at 90^th, 95^th, and 97^th percentiles. An elevated risk of LGA was found for ≥90^th percentile (OR 1.12, 95% CI: 1.03, 1.22), with steadily increased risk for ≥95^th percentile (OR 1.26, 95% CI: 1.13, 1.41), and ≥97^th percentile (OR 1.35, 95% CI: 1.17, 1.55). The estimated effect on LGA remained relatively unchanged even after adjustment. Agricultural household status was also associated with increased risk of postterm birth (aOR 1.20, 95% CI: 1.09, 1.33), and a low APGAR score at 5 minutes <7 (aOR 1.39, 95% CI: 1.07, 1.81). Conversely, LBW and PTB were inversely associated with agricultural household status (aOR: 0.85, 95% CI: 0.76, 0.96 and aOR: 0.82, 95% CI: 0.74, 0.92, respectively), and consistent with our findings that presence of a farmworker in the household is associated with larger birth size.

Table 4 shows the effect when the agricultural worker was the mother or father. Firstly, the majority of the agricultural homes had a father as a farmworker. Overall, the findings remained statistically significant, except for low APGAR score. Associations with low APGAR score were much stronger for neonates born to mothers who were agricultural workers versus fathers (aOR 3.72 for mothers, aOR 1.34 for fathers), although these ORs were statistically significant for both fathers and mothers. Associations with postterm birth were similar by whether mothers or fathers were agricultural workers (aOR 1.39 and 1.20, respectively).

Sensitivity analyses that included gestational weight gain and smoking during pregnancy as covariates did not change the magnitude of the effect estimates by more than 10% (see Table, Supplemental Digital Content 1). Similarly, when we excluded infants with a mother with chronic hypertension, the estimates of effects remained consistent (see Table, Supplemental Digital Content 2). Likewise, there was no change in effect measures after exclusion of births from parents who worked as farmers, growers, horticulturalists, and/or supervisors (n=372). Stratification by fetal sex showed a higher risk of LGA above the 90^th percentile among boys (OR: 1.24, 95% CI: 1.10, 1.40) compared to girls (OR: 1.04, 95% CI: 0.92, 1.17) with agricultural household exposure (p-interaction=0.04), although we did not observe any interaction for the other outcomes.

Abnormalities of fetal overgrowth are attributed to increased glucose in mothers that induce hyperinsulinemia, resulting in adipogenesis and increased oxygen demands upon the placenta.⁹ The effect of maternal overnutrition and excessive gestational weight gain increase the likelihood of women delivering LGA and macrosomic newborns. Moreover, fetal overnutrition leads to overgrowth, complicating labor and delivery for mothers, increasing the risk of neonatal mortality or birth trauma like respiratory distress and shoulder dystocia.⁸ Additionally, increased birth weight and conditions of overgrowth are associated with risk of obesity later on in childhood and adulthood.⁸ Women of reproductive age with co-morbidities like diabetes, hypertension, obesity and have had a prior macrosomic or postterm birth are at an increased risk of fetal overgrowth during gestation.

Results from environmental health studies utilizing maternal self-report on agricultural occupation and birth indicators, birth weight and gestational age (GA), have been inconsistent. We found positive associations with conditions of excessive fetal overgrowth and agricultural household status. Overall, the prevalence of macrosomia in the general US population is 7.8% of live births, similar to findings in our non-farmworker population.²⁵ Given the adverse short-term and long-term outcomes for both the mother and child associated with macrosomia, some researchers have hypothesized the cost-effectiveness of labor induction with suspected macrosomia to prevent burdensome complications.²⁶ In our study, agricultural workers were 15%, 38%, and 48% more at risk of delivering a macrosomic neonate at ≥4,000g, ≥4,500g, and ≥5,000g, respectively, than their non-farm work counterparts. In a similar manner, agricultural workers had higher odds ranging from 14 to 33% for delivering LGA neonates delivered at percentiles 90^th, 95^th, and 97^th. Our observed associations are similar to findings from a study of Polish mothers who reported working in the fields during 1^st and 2^nd trimesters; these mothers delivered larger newborns compared to non-farming mothers. This positive association is also supported by findings among neonates born to agricultural workers in Denmark.^27,28 Our findings that longer GA is also attributed to parental agricultural work aligns with a study of greenhouse workers.²⁹ Similarly, in a study of paternal occupational exposure, higher birth weight of neonates are observed among father agricultural workers compared to non-agricultural fathers.³⁰ In addition, studies of associations of fetal growth restriction indicators (i.e. LBW, PTB, SGA) among father agricultural workers show inverse associations.^22,31 These findings suggest a plausible role for epigenetic influences of father’s sperm, although in our study it may also represent environmental exposures from track-in into the home.^3,4

We also found lower APGAR scores for neonates born into agricultural households, which concurs with prior studies showing associations with exposure to manual labor.³² APGAR scores are an important marker for neurodevelopment in childhood and adulthood, and future research of workers exposed to agrochemicals will help characterize the risks posed to the general population of pregnant individuals. Preexisting literature suggests that the primary contributors of fetal overgrowth during gestation are related to exposures to heavy job strain and environmental endocrine disruptors, such as pesticides.

While the effect of agricultural job tasks and working conditions (i.e., prolonged standing, repetitive lifting, long shifts) on birth outcomes have not been fully examined, one study found that mothers with heavy lifting/high job strain, increased their risk of delivering a LGA neonate compared to mothers with low physical/low stress jobs.³³ Heavy manual labor is a known risk factor for pregnancy complications, and hypothesized to alter maternal plasma volume and blood flow to the placenta.⁷ Additionally, occupational stress has been linked to increased glucose levels that can give rise to diabetes, a risk factor for macrosomia and LGA.³⁴ Working women, but in particular Latinas, are vulnerable to unsafe working conditions due to their low social status typically associated with their immigration status. Further, agricultural workers may experience language and cultural barriers that impede access to quality prenatal care. Metabolic effects of heavy job labor and stress may also be exacerbated by exposure to pesticides, which have been positively associated with type 2 diabetes among agricultural workers.³⁵

Pesticides have various toxicological mechanisms due to their different active ingredients, making it particularly challenging to examine their effects on fetal growth.³⁶ However, many pesticides are endocrine disruptors that can impair hormonal function during pregnancy, as well as growth and development of the fetus.² Chronic exposure to pesticides during pregnancy may cause deleterious harm; as pesticides are metabolized by the mother to cross the placental barrier.³⁷ Although we do not have information specifically on which pesticides these farmworkers were exposed to during their work, we can presume that at least some of these exposures were organophosphate pesticides, carbamate pesticides, and/or pyrethroid pesticides, since OPs and pyrethroids were two of the most commonly used classes of insecticides during this decade.

OPs may regulate fetal growth through disruption of choline and acetylcholine receptors, which are critical components of fetal growth and development.³⁸ In our study, positive findings of fetal overgrowth and farmworker household are consistent with studies measuring biomarkers of prenatal OPs exposures in longitudinal birth cohort studies. In an agricultural community in California, non-significant trends in birth weight with maternal urinary dialkyl-phosphate metabolites (DAPs, a general metabolite for most OPs) were observed, with the highest increase of 52 grams for every 10-fold increase in diethyl-phosphates metabolites (DEPs).³⁹ However, a pooled analysis (N=1,169) from four U.S. birth cohorts found no significant association between OPs and birth weight.⁴⁰ Conversely, the lesser studied carbamate pesticides, which have a similar mode of actions as OPs, have been inversely associated with birth weight.^41,42

Pyrethroids may disrupt fetal growth and functioning through disruption of voltage gated sodium and ion channels, which are critical for neurodevelopment.⁴³ They may also disrupt critical signaling processes in labor and delivery, as the voltage-gated ion channels targeted by pyrethroids play critical roles in initiating and maintain labor.⁴⁴ In Asia, pyrethroids have been implicated^45,46, but inverse associations have also been noted.^27,47–49 Studies of other types of pesticides and pesticide mixtures are largely contradictory.^20,27,45,50 More epidemiological research is needed to determine the co-exposures of chemicals that may impair fetal growth among pregnant populations.

Previous research from U.S. studies which examined GA have reported mostly inverse associations with pesticides, namely, OPs.^39,51–53 Pyrethroids were implicated for longer GA and decreased risk of SGA and PTB in a Chinese population.⁴⁵ Evidence to-date is largely speculative, which may be the result of GA acting as a mediator in the relationship between fetal size and prenatal environmental exposures. Studies of agricultural pesticides and adverse birth outcomes have likely yielded discordant results due to the disparate study designs, detection of parent compounds/metabolites, covariates, timing of exposures, PON1 enzymatic activity, dynamic employment statuses, background community-level exposures, and geographic variations in climate and pesticide levels around homes during pregnancy.

Data from Arizona birth certificates over an 8-year period allowed investigation of potential risks of pregnancy complications from agricultural households. This allowed us to identify mothers and their partners who worked in agriculture and, therefore, to examine household exposures. Few studies have considered parental employment status and risk of adverse pregnancy outcomes.^7,33 Next, this study could be considered a type of mixtures analysis. Traditionally, study design methods were limited to one environmental exposure to one pregnancy outcome. Misclassification could occur when a single chemical class of compounds is examined, because in the real world, joint exposures of industrial and consumer by-products could occur in-tandem. One study of non-persistent chemical exposure mixtures, including phthalates, bisphenols, and pesticides found that women with the highest levels of chemical mixtures exposures, compared to the lowest had greatest difference in fetal weight gain at delivery.⁵⁰ By utilizing agriculture household status as a proxy for various exposures, we invariably account for multiple occupational exposures.

Several potential limitations are due to the use of birth certificate data. Misclassification of exposure from self-reported classification of agricultural occupation may occur; however, research has shown good reliability compared to maternal interviews, suggesting that this bias is relatively minor.⁵⁴ In this population, we assume minimal nondifferential misclassification due to high agreement, but also higher specificity and low sensitivity, so that fewer individuals are missing from the detected group of agricultural workers. Exposure misclassification may also occur as the result of workers shifting work status during the pregnancy from potentially higher exposure levels in early/mid-pregnancy to lower exposure status for the remainder of gestation and at the time of delivery. Birth records could also be subject to outcome misclassification, although other studies have shown that birth certificates performed well for GA and birth weight, respectively 98.9% and 98.6% in agreement.⁵⁵ While there may be some discrepancy between birth certificate and medical records, records are also prone to missing data and recording errors and, unlike birth certificates, are not standardized across clinics/hospitals. Finally, another limitation could be the effect of unmeasured environmental confounders (i.e., pesticides). Additionally, we could not account for several other environmental factors, such as housing quality, which are known to be sources of pesticides exposures for women and children. Since environmental information is not included in birth certificates, we used parental farm work as a proxy for mixtures of environmental exposures. Other potential confounding variables like maternal diet, alcohol and/or drug use were not included in the final models, because of either high rates of missingness, or evidence that they were unreliable.

This study of births in Arizona is one of the few studies that considers parental exposures to agricultural work and adverse pregnancy outcomes. Agriculture and farm field laborers belong to one of the most dangerous and hazardous industries with limited occupational and legal protections to enforce workplace safety standards.⁵⁶ In the U.S., the agricultural workforce comprises a vulnerable group of low-wage, immigrant workers and predominately foreign-born Latinos, making it challenging for research to address negative health outcomes. Future epidemiological studies should consider joint exposures of social and chemical stressors on fetal growth in under-resourced communities of color where environmental hazards are often most ubiquitous.