The SAWASDEE study is based on a longitudinal birth cohort of women residing in northern Thailand who are agricultural workers or live on a working farm. The exposures of interest are pyrethroids and OPs, while the outcomes of interest are birth outcome and child neurodevelopment. All study procedures were reviewed and approved by the institutional review board at Emory University (with Rutgers reliance) and the ethics review committee at the Research Institute for Health Sciences, Chiang Mai University (with Chulalongkorn reliance). Informed consent was obtained from all study participants prior to enrollment.

Recruitment occurred in Chiang Mai Province in northern Thailand because of its robust agricultural sector and its generalizability to other low-/middle-income countries (LMICs) that are similarly reliant on agriculture. The primary crops in Chiang Mai include rice, tangerines, longans, lychee, and ornamental flowers, with their use of pesticides varying greatly with the crop grown. For example, fruits trees require a larger application amount in a shorter application period. Conversely, rice crops require insecticide application more regularly during the growing season but at lower quantities. These 2 types of pesticide application scenarios (ie, high application, short duration, and low application, longer duration) can be easily studied in Chiang Mai Province by focusing on 2 of its distinct districts: (1) Chom Thong and (2) Fang. Chom Thong and Fang are both agricultural communities located about 60 and 150 km, respectively, outside of Chiang Mai City. Chom Thong crops include longans, cut flowers, vegetables, and rice, while in Fang, located in the mountainous region near the Myanmar border, tangerines are the predominant crop. The differences in crops grown in the corresponding districts create a sort of natural experiment of exposure, allowing us to evaluate the differences in these scenarios.

Prior to recruitment, we developed and culturally validated, where appropriate, a detailed set of standard operating procedures (SOPs) and questionnaires. SOPs were developed for sample identification numbers and label generation, biospecimen collection and processing, laboratory testing, and quality assurance and control (QA/QC) procedures, including field blank sample collection, bench-level QC and blank samples, sample compositing, sample randomization for analysis, and neurological testing. In addition, to facilitate sample processing and efficient transfer of samples to Chiang Mai University, we scoped and rented apartments that we outfitted as satellite laboratories we called “SAWASDEE houses” located in each of the districts.

Recruitment began in July 2017 and was completed in June 2019. Women were recruited at community and district hospitals in both districts by trained study nurses. Nurses administered an initial screening questionnaire to determine eligibility of the pregnant women. Women were eligible for participation if they (1) were an agricultural worker or lived within 50 m of an agricultural field, (2) had a Thai identification card permitting hospital and antenatal clinic access, (3) resided in their regional district for ≥6 months and planned residence at least 3 years after delivery, (4) spoke Thai at home, (5) were in good general health (ie, no major medical conditions, such as hypertension, diabetes, thyroid disease, or HIV), (6) consumed fewer than 2 alcoholic beverages per day and did not use illegal drugs [13], (7) and were at <16 weeks’ gestation. Expectant mothers with nonsingleton pregnancies or major pregnancy complications that could affect fetal growth and development were excluded from further participation at the time of diagnosis.

We screened 1289 women, and of those, 398 (30.88%) met our enrollment criteria and of which 394 (98.9%) were enrolled. Of these 394 participants, 322 (81.7%) completed all data collection (Figure 1) through their child’s neurological visit at 1 month. Of these 322 participants, 217 (67.4%) were from Chom Thong and 105 (32.6%) were from Fang, with the proportion of total participants reflecting their relative population sizes. Most of the participants who were not retained in the study were excluded due to pregnancy complications (eg, miscarriage, blighted ovum) or because they moved away from the study area.

Trained research assistants (RAs) and nurses administered each questionnaire in Thai. These questionnaires included a recruitment survey to determine eligibility (screening questionnaire); an intake questionnaire to collect baseline descriptive statistics; an extensive exposure questionnaire administered during early, middle, and late pregnancy that roughly corresponded to trimesters; and a Knowledge, Attitude, and Practices (KAP) survey of pesticides, collected during the first and third trimesters. The exposure questionnaires, which were validated and used in our pilot study [14], included questions on work-related tasks, household and occupational use of pesticides, personal habits of parents, food consumption, household characteristics and cleanliness, medical histories, maternal nutritional status during pregnancy, and demographics. These questionnaires include detailed information about the population characteristics, including information about medical conditions, medications, weight gain during pregnancy, education, family assets, supplements taken during pregnancy, parity, and more. By administering the exposure questionnaire 3 times, we were able to capture any changes that occurred that might have altered exposures. The KAP survey was used to evaluate the participants’ understanding of the toxicity and potential effects of pesticides and their routine practice related to pesticides and how those practices changed as pregnancy progressed. After delivery, an additional 9 questionnaires were completed over the course of 36 months to ascertain developmental milestones, maternal intelligence quotient (IQ), family assets and the home environment, and anthropometrics of the children to evaluate their nutritional status.

The following information was extracted from medical records: early pregnancy weight and weight gain, parity, past obstetric history, planned pregnancy, smoking and alcohol consumption, marital status, occupation, history of illness, and prenatal visit history, which included gestational age at first visit, subsequent visit dates, doctor-assessed nutritional status, and compliance with micronutrient supplementation regimens. These data were updated at each of the visits to track a variety of indicators, including maternal nutritional status (using weight gain as a surrogate) and significant stressors that could complicate the exposure-effect evaluation. Data extracted from participant birth records included sex, birth weight, length, gestational age, and head circumference; appearance, pulse, grimace, activity, and respiration (APGAR) scores; type and route of delivery; anesthesia; and information about maternal and neonatal complications.

Biospecimens collected from enrolled participants are shown in Table 1. Most of these samples were processed and archived for planned future analysis or novel research questions.

Blood, serum, urine, and hair were used for exposure assessment, with urine being the primary matrix used. Blood was used for susceptibility biomarker analysis (ie, paraoxonase 1 [PON1] phenotyping), and placenta tissue was used for transcriptomics analysis.

To provide trimester-specific exposure measures during pregnancy, while still keeping the study costs reasonable, we created trimester-specific urine composites from each participant but still retained the original discrete samples for later use. DAPs, TCPY, and 3PBA were measured in these maternal composited urine samples (Table 2) using established analytical methods [15-17]. PON1 enzyme activity, a measure of OP insecticide susceptibility, was measured in maternal serum using previously established procedures [14]. Other samples were archived for future use.

Extensive neurodevelopment examination of each child was conducted by trained neuropsychologists at the following time points: 1, 4, 7, 12, 18, 24, and 36 months (Table 3).

Assessment at these intervals represents a significant departure from previous studies of pesticide exposure and will allow examination of how pesticides may alter the neurodevelopmental trajectory of children. For tests administered at more than 1 developmental period, age-appropriate adjustments were made to increase the difficulty of the test. Infant and child responses were either scored in real time in the case of visual habituation tests or scored from digital recordings when testing was complete. Interrater reliabilities were conducted monthly for all tests using different videos each month and independent scoring by psychologists to maintain reliability among testers. All tests were routinely recorded to allow recheck procedures and scoring. The neurodevelopmental measures used in our study included those traditionally used in other birth cohort studies (eg, Bayley Scales of Infant and Toddler Development at age 3 years) but also included tests of visual habituation at 4 and 7 months as indicators of visual attention and information processing, followed by tests of infant memory and executive function (Tables 1 and 3). The NNNS was used within the first week of delivery and is a comprehensive assessment of both neurological integrity and behavioral function. Additionally, tests of emotion regulation and inhibitory control were administered to determine how not only cognitive skills but also emotional development may be altered by exposures to pesticides. This more comprehensive battery of tests will enhance our ability to determine whether and at what developmental period prenatal pesticide exposure alters the building blocks of later cognitive, motor, and emotional development.

The Home Observation for the Measurement of the Environment (HOME) is a structured interview conducted with the primary caretaker and child in the home of each participant to observe and assess the physical environment, family structure/relationships, and learning environment for the child [18]. HOME was administered at 12 and 24 months. HOME has been extensively used across the world, with adaptations for different cultures, and is predictive of child cognitive development [19]. Mothers were administered the Test of Nonverbal Intelligence (TONI)-I as a nonverbal measure of intelligence in recognition of the effects the mother’s education and cognitive abilities have on child neurodevelopment. The Depression Anxiety Stress Scale 21 (DASS-21) is a measure of maternal depression and anxiety that has been used in cross-cultural settings, including Asia [20]. The DASS-21 was administered at the first antenatal visit (1 week after birth) and at every subsequent child assessment visit to assess the known effects of maternal depression and anxiety on development. Although we asked families to estimate their monthly income, in low- and middle-income agricultural settings, monthly income estimates are difficult to determine and may not fully capture the economic health of a family. Therefore, based on the Economic and Social Research Council (ESRC) Research Group on Wellbeing in Developing Countries [21], we developed a family assets questionnaire to augment income estimates [22].

In total, 1289 pregnant women were administered the recruitment-screening questionnaire. The primary reasons screened individuals were not eligible were as follows: (1) their gestational age was more than 16 weeks (n=266, 20.64%) or (2) they were agricultural workers or lived within 50 m of an agricultural field (n=603, 46.78%). Of those individuals eligible for the study (n=398, 30.88%), 394 (98.9%) women provided consent to join the study and were enrolled (Figure 1). Among the participants, 322 (81.7%) completed the study through the first neurodevelopmental examination of their child, the NNNS.

Of the 322 (Chom Thong n=216, 67.1%; Fang n=106, 32.9%) mother-child pairs with a completed NNNS, the mean age of SAWASDEE study participants at enrollment was 25 years (SD 5.3); see Table 5. The mean gestational age at enrollment was 9.9 weeks (SD 2.6 weeks). In Chom Thong, 47 of 216 (21.8%) and 165 of 216 (76.4%) women reported being legally married or living as married, respectively, compared to 1 of 106 (0.9%) and 102 of 106 (96.2%) women in Fang. The mean gestational age at birth was 38.5 weeks (SD 1.2 weeks), with gestational age determined by ultrasound or last menstrual period differing by participant site (n=105/216, 48.6%, and n=106/106, 100%, determined by ultrasound at Chom Thong and Fang, respectively). The mean infant birth weight was 3 kg (range 1.7-4.2 kg) and was not statistically significant (P<.05) between participants from Chom Thong and Fang.

The median and range of exposure measures and PON1 activity categorizations are shown in Table 6. Overall, exposure to OPs was higher in Fang than in Chom Thong in keeping with their crop application rates. Levels of pyrethroids were similar across sites. Overall, about 29 (9%), 126 (39.1%), and 167 (51.8%) of 322 participants had low, normal, and high PON1 activity, with more participants in Fang having normal activity than in Chom Thong.

The SAWASDEE study was successfully established to fill critical gaps in the literature by measuring exposure multiple times during each trimester of pregnancy, while tracking neurodevelopmental trajectories in infancy and early childhood. By establishing the association between exposure biomarkers during early-to-late pregnancy and these neurodevelopmental trajectories, we will be able to examine windows of increased susceptibility to pesticide exposure. This information may result in specification of vulnerability, resulting in recommendations for reduction in adverse neurodevelopmental outcomes.

Our enrollment and participant rates were notable, provided the low population density and complexities of travel in these regions. The terrain is mountainous and winding; participants had to often rely on public transportation to make the study visits. We tried to align our study visits as closely as possible to their antenatal visits for convenience and reimbursed participants for travel. However, in some of the worst cases, participants traveled for up to 4 hours to attend these visits. Their loyalty to our study was largely a result of the close relationship our study nurses and RAs developed with the participants and the immense trust our participants had in them. Although we believe that the relationships that were established between field staff and study participants were essential in preserving study retention, it is also possible there were behavioral changes in pesticide practices over time due to unintentional influence by field staff or involvement in the study. Analysis of the KAP questionnaires from our pilot study revealed that there was an increase in pesticide-related knowledge among pregnant women and that influenced changes in pesticide practices [24]. In this study, we did observe that many women stop working in agricultural-related fields late in pregnancy; however, we were unable to discern whether this is a result of the physical challenges associated with pregnancy, especially in the third trimester of pregnancy; a natural shift in agricultural tasks required; or behavioral change from participating in our study. We hope to disentangle the reason(s) for a reduction in agricultural work late in pregnancy in future analyses using the longitudinal KAP data collected in our study.

Our study is not without limitations. We may have some differential bias in gestational age among sites, which we described above. There is potential for exposure misclassification because of the contribution of preformed environmental metabolites to the measured urinary metabolite levels. However, given that our population is experiencing occupational exposures that occur after known pesticide application events, we expect this contribution to be smaller than if we were evaluating background dietary exposures. We also did not measure all pesticides to which participants may have been exposed. Although we have basic dietary information for mothers during pregnancy, we could not conduct a thorough assessment of dietary information, such as a food frequency questionnaire, largely because of funding issues. In addition, some of our neurological tests have no known cultural norms for comparison, but our pilot studies have provided insight into how this can be addressed in our data analyses [14].

Despite our study’s limitations, this is still the first study to evaluate these neurodevelopmental outcomes in Thailand and to provide refined exposure measurements during pregnancy.

Our study has several strengths. We have time-resolved and highly sensitive exposure measures from all trimesters of pregnancy in our participants. This includes 2 distinct exposure scenarios: long-term, lower-dose pesticide exposure, and high-dose, intermittent pesticide exposure. The robustness of exposure and outcome data will allow investigators to evaluate numerous scientific questions and ultimately improve the generalizability of associations between pesticide exposure and neurodevelopmental outcomes. The majority of previous studies have been conducted using cohorts from high-income countries and examination in LMICs is lacking. The success of our study recruitment and retention is largely attributed to the strong community-centered nature and the respect shown to the field research staff by study participants. We found this to be especially true within the Thai culture—individuals have a high level of respect for health care practitioners. The study also includes highly refined neurodevelopment assessment with sensitive measures of early neurointegrity that measure the basis of later cognitive and motor development. Previous studies have used clinical measures that are relatively less sensitive to subtle and early adverse effects. We can evaluate windows of vulnerability as they relate to specific neurodevelopmental processes and the trajectory of development. In addition, this study assesses not only cognitive and motor skills but also inhibitory control and emotion regulation, not previously studied with behavioral measures. The NNNS, in addition to the other neurodevelopmental assessments included in the study, creates a robust database of neurodevelopment, which can be used to evaluate several scientific questions related to neurodevelopment, including evaluating trajectories of neurodevelopment. The prospective longitudinal data in our study present a valuable resource for evaluating many scientific questions. The SAWASDEE study offers potential collaborators an opportunity to work with robust estimates of exposure and neurodevelopment. The number of supplemental studies that are part of the SAWASDEE study continues to increase, and thus, the database continues to grow, offering investigators an opportunity to test several hypotheses.

The SAWASDEE birth cohort study provides information essential for risk assessment paradigms addressing the risk of prenatal pesticide exposure and neurodevelopment. The cohort is unique in that it is designed with highly refined exposure and neurodevelopmental data with the ability to determine neurodevelopmental trajectories early in infancy with the hope of early intervention. Successful enrollment followed by high rates of retention and participation in the study provides a unique opportunity for evaluating maternal pesticide exposure and child neurodevelopmental outcomes.