Chlorpyrifos and other pesticide exposure and suspected developmental delay in children aged under 5 years: a case-control study in Phitsanulok, Thailand

Background: Developmental delay among children under 5 years of age is a serious global public health problem and much research has been carried out to find potential causes. Pesticides - especially organophosphates - are suspected to be one of the main causes of the problem. This study aimed to investigate the association between pesticide use by the mother during pregnancy and preschool children development using a case-control study. Methods: Data on prenatal and postnatal pesticide exposure of 442 children with suspected developmental delay, and 413 controls with normal development were included for analysis. The children were matched for gender, age, and residency. Data on pesticide exposure were collected via interview with the mother, and data on pregnancy outcomes abstracted from hospital records. Results: Chlorpyrifos exposure significantly increased the risk of developmental delay with an odds ratio (OR) of 3.71 (95% CI 1.03-13.36) for ever use of the pesticide, and an OR of 5.92 (95% CI 1.01-34.68) for postnatal exposure (p <0.05). Some other pesticides also had a positive association with developmental delay but none were statistically significant (p <0.05). Those pesticides were insecticide, fungicide, herbicide, and molluscicide. Individual pesticides with a positive association were glyphosate, paraquat, butachlor, methyl parathion (pholidon), savin, methomyl, endosulfan, carbosulfan, methamidophos, monochrotofos, mancozeb, and bordeaumixture. Conclusions: This case-control study found that chlorpyrifos and some other pesticides exposure during pregnancy were positively associated with developmental delay in children aged under 5 years. Further research should be conducted to better understand this potential effects of pesticides on child neurodevelopment, and the public - especially those who plan to have families - should be informed.


Invited Reviewers
23 Dec 2020 report report

Introduction
Developmental delay in young children is a global public health concern. A study of 35 low-and middle-income countries reported that one in every three children below five years of age fails to reach their developmental potential 1 . In Thailand, a national survey by the Ministry of Public Health reported that approximately 15% of children aged under 5 years are suspected to have a developmental delay (SDD) 2 . In addition to stunting, inadequate cognitive stimulation, iodine and iron deficiency, malaria, intrauterine growth restrictions, maternal depression, exposure to violence 3 , exposure to environmental toxicants including phthalates, bisphenol A, flame retardants, polycyclic aromatic hydrocarbon (PAHs), gas cooking 4 , and heavy metals 3 .
Pesticides of the acetylcholinesterase inhibitor group are another set of compounds suspected to affect neurodevelopment. In laboratory studies, this type of pesticide has been found to affect neuron cell and synaptic functions 5 . Young children are also at a higher risk of pesticide effects because their bodies are not yet fully developed, and they also have a higher chance of exposure to environmental pesticides from engagement in high-risk behaviors, e.g. crawling on the floor, object-to-mouth behaviors, and playing with items found in the environment 6 . A recent literature review indicated that 45 out of a total of 50 articles found a positive association between delayed neurodevelopment in young children and OP exposure 7 . The neurological and behavioral developmental outcomes induced by pesticides include slower neonatal reflexes, delayed psychomotor and mental development 8 , attention deficit 9 , lower IQ 10 , and autism spectrum disorder (ASD) 11,12 .
Regarding individual pesticides, chlorpyrifos (CPF) is the only OP pesticide that has been extensively studied, with most studies finding a positive association between exposure to CPF and child developmental delay. In a study among inter-city minority communities in New York City, researchers reported a positive association between levels of CPF in umbilical cord plasma and neurodevelopmental delay 13 . Studies have also linked CPF exposure to poorer outcomes in working memory, visual motor coordination, color discrimination 14 , and verbal comprehension 15 , in addition to vision and hearing loss 16 , and lower IQ 17 . There is limited evidence that CPF exposure might also relate to ASD 18 . On the other hand, a study of 2-year-old Mexican-American children found no association between cognitive function and CPF exposure 19 , meaning this link is not yet conclusive.
Currently, in Thailand, the effects of CPF on child development have not been adequately studied. The objective of this case-control study was to analyze the association between suspected developmental delay (SDD) in children aged under 5 years living in Phitsanulok province, Thailand, and the use of CPF and other compounds during pregnancy. The results may be useful for SDD prevention, and for comparison to other similar studies in this field.

Study area
Phitsanulok province is in lower northern Thailand, located 370 km from Bangkok. It is a midsize province of 4,176 square miles, with nine districts, and a population of 866,891 people (density = 200 people per square mile). The capital city of the province is Muang district.

Study participants
This study used a case-control design. Children diagnosed with suspected developmental delay (SDD) (cases) were compared with normal children (controls) with respect to pesticides exposure of the mother during pregnancy. Both cases and controls were children aged under 5 years who had participated in the National Child Developmental Screening Program. In Thailand, every child is screened for development progress at the ages of 9, 18, 30, and 42 months using the Developmental and Surveillance and Promotion Manual (DSPM) which modified from Denver Development Screening Test II (DDST-II). The screening is carried out by a trained nurse or health personnel at a health promoting hospital. In accordance with the DSPM manual, the children are evaluated in five skills, namely 1) gross motor skills), 2) fine motor skills, 3) receptive language skills, 4) expressive language skills, and 5) personal and social skills. If a child fails one or more of these skills, they are classified as having SDD. Children classified as having SDD were the target population in this study and were randomly selected to take part. The controls were children who attended the same hospital for the screening program but passed all five skills and were therefore classified as having normal development. The case and control groups were matched for gender, age, and area residence at assessment. Children with congenital anomalies or head trauma were excluded from the study.

Sampling and sample size
Participants were children who participated in the screening program in selected local hospitals in the Bang Rakam and Muang districts of Phitsanulok province, Thailand. These two of the nine districts in the province were purposively selected to represent a rural area (Bang Rakam district) and an urban area (Muang district) of the province. A total of 15 out of 21 local hospitals in the Bang Rakam district, and 10 out of 30 hospitals in the Muang district were randomly chosen using a simple lottery method. The mothers of every child with SDD who met the inclusion criteria from the selected hospitals were invited to take part in the study at their appointment. For each case, a child with normal development was randomly selected from

Amendments from Version 4
In the discussion section, more information on study limitations has been provided. Recall and selection bias was further discussed.
Any further responses from the reviewers can be found at the end of the article REVISED the hospital database matching gender, age, and area of residence.
The minimum sample size was calculated to be 816 (408 cases and 408 controls) using OpenEpi online using the following assumptions: confident interval = 95%, power of detection = 80%, ratio of case to control = 1:1, proportion of control with exposure = 40, odds ratio = 1.5 20 .
Questionnaire and data collection Data on pesticide use and exposure during pregnancy was collected from the child's mother using a constructed questionnaire (provided as Extended data in English) 21 . Besides demographic data, the children's mothers were asked "yes" or "no" questions concerning prenatal and postnatal use of pesticides. The exposure period was classified as "ever" for any prenatal and/or postnatal exposure to pesticides, "prenatal" for prenatal exposure to pesticides, and "postnatal" for postnatal exposure to pesticides. Pesticides were categorized into insecticides, herbicides, fungicides, rodenticides, and molluscicides. Exposure data for 14 individual compounds that are commonly used in Thailand and around the world were also collected. There were also questions for potential confounding factors such as occupation, monthly income, education, cigarette smoking, and alcohol use of the mother. Data on health status and pregnancy outcomes including delivery method, gestation, birth order, birth weight, and breast-feeding history, were retrieved from the hospitals' medical records. Data collection took place in the participants' homes, and was conducted between January and May, 2019. Data were collected by 60 village health volunteers who were trained to use the questionnaire and conduct the interviews.
The questionnaire was constructed by literature reviewed. The content validity of the questionnaire was tested by three experts in pediatric, obstetrics and gynecology and family medicine, and occupational health nursing. The index of Item Objective Congruence (IOC) was between 0.67-1.00. The questions were also tested for sequencing and understanding with a group of 30 women with similar characteristics of the intended participants.

Data analysis
Demographic characteristics were analyzed with descriptive statistics and the results presented as frequency, percentage, mean, and standard deviation. Differences between groups were compared via t-test for continuous variables, and chi-square test for categorical data. The association between developmental delay and pesticide exposure was analyzed using multivariable logistic regression with odds ratios (OR) and a 95 percent confidence interval (CI) adjusted for mother age when pregnant (continuous), education (no school, primary school, secondary school, college degree), occupation (farmer, own business, civil servant, employee [formal], employee [general work], housewife, retired, unemployed), income (<5000 baht, 5000-9999, 10000-14999, 15000 or more), chronic disease (yes, no), alcohol consumption (yes, no), gestation (<37 weeks, 37 or more weeks), birth order (1, 2, 3 or more), delivery method (vaginal delivery, caesarean section, assisted delivery), baby weight (<2500 grams, 2500 grams or more), and breast-feeding (yes, no). These variables were those found from the literature to be potentially confounding factors, and those with significant differences between case and control groups. For variables on environmental pesticide exposure, they were not included in the model because they were either not significantly associated with developmental delay (Table 1), or strongly associated with pesticide exposure use (data not presented). Data was analyzed using IBM SPSS (version 26) software. Statistical significance was set at p <0.05 (2-tailed test).

Ethical considerations
Ethical approval for this study was obtained from Naresuan University Institutional Review Board (Approval number 448/2019). Written informed consent to participate in the study and for attaining of their clinical details were obtained from the parents of the patients before data collection.

Demographic data
From the dataset of 858 individuals, 855 records (413 cases, 442 controls) were used in the data analysis. Three records were not included for analysis because important information such as gender and age, were missing. The overall participation rate was 83.8% (86.7% for the case group, and 81.0% for the control). Demographic data of the participants is shown in Table 1 and in the Underlying data 22 . Most of the mothers were in the youngest age group with an average age of about 25 years. Most of them finished secondary school and had a monthly income of about 10,000 Thai Baht (300 USD) which is the minimum wage for Thailand. Only about 10% of them were farmers, and therefore reported using pesticides. There was a significantly higher proportion of mothers working as private employees, housewives, and civil servants. Most of the participants were healthy and had never drunk alcohol. One participant reported smoking cigarettes, and thus, was excluded from the data analysis. Data from the child's medical records revealed that most of them, with the same proportion of case to control were born with spontaneous vaginal delivery (n=303, 69.5% and n=274, 67.5%, respectively). Compared to the control group, there was a higher proportion of cases born preterm and being the second or third child of the family. There was also a significant difference in birth weight and breastfeeding between groups. Although a higher percentage of cases had a birth weight of below 2,500 grams (n=64, 14.5% vs n=31, 7.6%, respectively), a lower percentage of them was not breast fed (n=25, 7.7% vs n=43, 12.8%, respectively).

Environmental pesticide exposure
Roughly half of the participants had lived in the community for more than 10 years. With an equal proportion in the case and control groups, about 35% of participants had a family member working on a farm, 20% often entered farmland, and 14% stored pesticides in the house (Table 1), yet around 70% lived within 1.0 km of farm land.

Association of pesticide use and developmental delay
There were only 47 (10.4%) case mothers and 46 (11.4%) control mothers who reported ever using any pesticides during pregnancy. Table 2 presented odds ratio of SDD by types of pesticides the children were exposed to during pregnancy. Types of pesticides and exposure periods that were positively associated with SDD were insecticides (PostN), fungicides (PreN), fungicides (PostN), herbicides (PostN), and molluscicides (PostN). However, none of these ORs were statistically significant (p <0.05).

Discussion
The results showed CPF exposure during pregnancy and childhood SDD, with an odds ratio of 3.71 (95% CI 1.03-13.36) for ever using the pesticide (either prenatal or postnatal exposure), 2.97 (95% CI 0.80-11.07) for prenatal exposure, and 5.92 (95% CI 1.01-34.68) for postnatal exposure (Table 3). Ever and postnatal exposure were found to be statistically significant. There was also a positive association, though not statistically  PreN)]. This is consistent with the literature: in an experimental study, CPF showed an ability to alter neuronal formation and structure in animal and human fetuses 23,24 . The synapse or neuronal junction, the site of transmission of nerve signals between two nerve cells, is perhaps a central target for neurodevelopmental susceptibility to pesticides. Since synapse plays a critical factor for the proper functioning of the neuro system, dysfunction of it, even subtle form, could lead to logic and psychiatric disorders, as well as subtler cognitive, psychomotor, and sensory defects 5 .
The results of epidemiological studies into SDD and pesticides have found a range of outcomes. In a study of Mexican American children aged 6-24 months, prenatal or child exposure to CPF was not associated with mental development, pervasive developmental disorder (a group of disorders characterized by delays in the development of socialization and communication skills), or behavioral problems 19 . However, several other studies have found a positive association between prenatal exposure to CPF and neurodevelopmental problems. A cohort study of three-year-old children from minority communities in New York City, USA, reported a high exposure group (CPF levels of >6.17 pg/g in the mother's plasma) to have a higher proportion of developmental delay, assessed by the psychomotor development index and the mental development index 13 . A more recent study in Costa Rica found 6-9-year-old children with higher CPF exposure to have several neurobehavioral problems, including poorer working memory, visual motor coordination, and color discrimination, as well as parentreported cognitive problems/inattention, oppositional disorder, and attention deficit hyperactivity disorder 14 . A recent study of 9-month-old Thai infants reported an association between prenatal exposure to CPF and a reduction in grating visual acuity (OR = 0.64, 95% CI -1.22 to 0.06) 16 .
One study has also linked prenatal exposure to CPF to lower IQ levels in children 17 . Similarly, a study among children aged 5.9-11.2 years linked prenatal exposure to CPF to brain anomalies 25 . This result has been replicated in a study of an adolescent group 26 . Neurotoxic deficits have also been associated with CPF exposure in high-exposure occupations 27 .
For groups of pesticides, most literature has focused on OPs due to their neurological toxic effects. These studies, usually using a cross-sectional or a cohort study design, have found positive correlations between OP metabolite in the mother's urine and neurodevelopmental problems in the child 7,8 . Studies in the USA have reported prenatal exposure to OPs to increase the risk of abnormal reflexes in neonatal children (OR = 2.24, 95% CI 1.55-3.24) 28 , and ADHD in male children at age five years (β = 1.3; 95% CI 0.4-2.1) 9 . Living in close proximity to agricultural areas using OPs and other pesticides during pregnancy has also been related to ASD and developmental delay 29 . A study in Taiwan using a case-control study design reported a dose-response relationship between OP metabolites in child urea and ADHD among children aged 4-15 years 11 . Studies have also linked OP, carbamate, and pyrethroid pesticide exposure to lower IQ 10,15,30 . A cohort study in Thailand also found lower motor and cognitive performance (using Bayley Scales of Infant and Toddler Development III [Bayley III]) among five-month old infants prenatally exposed to OP 31 .
For other pesticides, data are limited. A study in Costa Rica reported a positive association of prenatal mancozeb exposure and lower social-emotional scores (β per 10-fold increase = -7.4 points [95% CI -15.2 to 0.4][measured by Bayley III]) in one-year-old infants 32 . This is consistent with the present study which also found an elevated risk of SDD among those exposed to mancozeb, with odds ratio of 1.87 (95% CI 0.59-5.93) for prenatal exposure, and OR of 3.97 (95% CI 0.60-26.38) for postnatal exposure (Table 3). A cohort study in Brittany, France, reported a negative association with neurocognitive development of 6-year-old children with prenatal exposure to pyrethroid 33 . A cohort study in 4-year-old children in Greece reported the association between prenatal exposure to the organochlorine compounds and neurodevelopmental effects 34 . A recent study in Indonesia reported a higher risk of small head circumference at birth to antenatal exposure to household non-OP pesticides (OR = -22.1 mm, 95% CI -36.5 to -7.6) 35 .
Overall, the literature is limited and inconsistent regarding the critical duration of pesticide exposure developmental effects. In the current study, both prenatal and postnatal exposure was related to an increased risk of SDD (Table 3), yet only postnatal exposure was significant. This might be the effects of a small sample size, recall bias, and imprecise exposure assessment via questionnaire. Moreover, most of the previous studies on the effects of CPF on neurodevelopmental effect focused only from prenatal exposure 13,16,17,25 . Only a few have examined the effects from both prenatal and postnatal exposure. One study that did 36 report a negative association of children's developmental quotients (DQ) -a numerical indicator of a child's growth to maturity across a range of psychosocial competencies -with prenatal exposure to OP but not with postnatal exposure. On the other hand, a cohort study in China found both prenatal and postnatal OP exposure increased the risk of developmental delay especially in the adaptive development (self-care skills), among two-year old boys 37 . In a laboratory study, CPF caused neurobehavioral impairment to a zebrafish when the exposure occurred in either the fertilization stage or embryonic stage 38 .
In the current study, there were some limitations that need to be mentioned. First, there was a smaller number of participants who had used pesticides during pregnancy than expected, which limits the power of association between the variables. In addition, it is difficult to study the effect of low-level pesticide exposure on growth and development because the outcomes can be affected by several factors including biological factors (e.g. stunting, infections, anemia, IUGR, preterm birth, birth weight, sex of the child, gestational age at delivery), psychosocial factors (e.g. inadequate cognitive stimulation, exposure to violence, maternal depression, household dysfunction), and maternal sociodemographic factors (e.g. poverty, low education, young age, smoking, drinking alcohol) 39 . The study also lacks data on maternal diet and intelligence quotient (IQ). These two factors have been reported to be associated with child development 40,41 . Other problems were exposure misclassification and selfselection bias which were likely to occur in this type of study. Recall bias will occur when the cases are aware of pesticides as a potential cause of SDD and can recall pesticide use better than the control groups. However, the information on the association between pesticides and child development is new and has not yet been publicized, especially in Thailand. Study participants were also likely to be exposed to pesticides in the environment, however, if this information bias occurs, it could only lower the strength of the reported association. Selection bias, will occur when the mothers of SDD children were more likely to volunteer than the controls. This problem may not affect the result much, as mentioned before the mother may not be aware of the association of pesticides and SDD, and it was found that the participation rate among the case group and the control group were similar.

Conclusion
This case-control study found a negative association between chlorpyrifos and some other pesticide exposure during pregnancy and preschool child development. This effect was found in both prenatal and postnatal exposure. More research, using a larger sample size, is still needed to confirm the study results and to identify more individual pesticides which may impact prenatal and postnatal growth and development of children. This potential effect of pesticides on child neurodevelopment should receive more attention by researchers, and the public, especially those who plan to have families, should be informed. adequately addressed in the paper.

Data availability
In the limitations section they state "However, if this did occur, it would have happened equally between groups. Exposure misclassification and self-selection bias were unlikely to occur because participants did not know or are aware of pesticides as a potential cause of SDD". This is simply untrue. One of the risks of retrospective exposure classification is that it is highly prone to misclassification. I think they need to acknowledge this in the limitations without dismissing it because they think it is non-differential bias.
Also, I think the lack of information on material diet and IQ could have affected the results. This should be stated in the limitations as well. Education is more of a determinant of wealth in the Thai populations I have evaluated and less related to IQ.
Lastly, I think the manuscript does require some editing for English to ensure verb and noun agreement. Eg., 'data are' instead of 'data is' because the word 'data' is the plural of 'datum'. I noticed a few other occurrences of verb/noun disagreement as well.

Competing Interests: No competing interests were disclosed.
Reviewer Expertise: Biomonitoring, maternal child health, birth cohorts, exposure assessment, environmental epidemiology

I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.
Author Response 20 Aug 2021 Chudchawal Juntarawijit, Naresuan University, Phitsanulok, Thailand

Comment:
While I think the reviewers did do a good job of responding to my critiques, I don't feel they were adequately addressed in the paper.
In the limitations section they state "However, if this did occur, it would have happened equally between groups. Exposure misclassification and self-selection bias were unlikely to occur because the participants did not know or are aware of pesticides as a potential cause of SDD". This is simply untrue. One of the risks of retrospective exposure classification is that it is highly prone to misclassification. I think they need to acknowledge this in the limitations without dismissing it because they think it is non-differential bias.

Response:
Thank you very much again for your prompt response and valued comments and suggestions.
Yes, we agree. Those statements have been deleted and some more information has been added to the limitations.
the name of the pesticides they'd used in the past. However, if this did occur, it would have happened equally between groups. It was also very likely that study participants were exposed to pesticides in the environment, and this information bias could lower the strength of the reported association." but this is simply untrue. Mothers with children with SDD are far more likely to either recall precisely the chemicals that were used or to believe that the chemicals they used were the cause of SDD. Thus, I think this study is highly susceptible to recall bias. Similar issues with selection bias in that those that have more pronounced SDD were likely to volunteer or than those we less pronounced SDD or controls and the controls may have other uncontrolled differences.
Was information on crop provided by the mothers? This may have allowed the authors the opportunity to validate some of their questions. 3.
The biggest limitation in this study is the retrospective exposure assessment via questionnaire. Ever recall using pesticides is likely quite imprecise and mothers may not wish to divulge that they were exposed to pesticides during pregnancy so as not to be "blamed" for their child's SDD. Exposure misclassification is not mentioned in the limitations yet all of the studies in the discussion including the one central Thai study used more objective measures of exposure (i.e., biological levels). When comparing the studies, the authors do not mention this. Similarly, the authors cannot derive any exposure magnitudes from this study which could obscure any underlying associations.

4.
Because the authors findings differ so much from other studies that prospectively evaluated exposure and disease, they should elaborate more on why this is likely. For example, most prospective studies have found that prenatal exposures mostly drove any developmental delays while early childhood exposures had much less of an impact which is directly opposite of what the authors found. In all likelihood, the differences are likely because of exposure misclassification because the other studies had better exposure assessments 1,2,3, 4,5 .

5.
It would be useful for the authors to let us know how their populations compare in demographic characteristics as the population in Northern Thailand or all of Thailand and how this may differ from other populations in the study.

If applicable, is the statistical analysis and its interpretation appropriate? Yes
Are all the source data underlying the results available to ensure full reproducibility? Yes

Are the conclusions drawn adequately supported by the results? Partly
Competing Interests: No competing interests were disclosed.
Reviewer Expertise: Pesticide exposure assessment, birth cohorts, maternal-child health, child neurodevelopment, north Thailand populations I confirm that I have read this submission and believe that I have an appropriate level of expertise to state that I do not consider it to be of an acceptable scientific standard, for reasons outlined above.

Comment:
This manuscript describes the findings of a case-control study in Northern Thailand that evaluates suspected developmental delay (SDD) in relation to pesticide exposure as assessed through a questionnaire. While the study has several strengths such as the large size, matching cases and controls on sex, age, area of residence and hospital of SDD testing, and the standardized clinical testing of SDD, the study also has some pretty significant limitations/weaknesses that are not appropriately addressed.

Response:
Thank you very much for reviewing the manuscript and providing insightful comments. Yes, we agree that those issues are important and need to be clarified.

Comment 1:
No mention of maternal IQ was provided. Many studies have recognized that this is a potential predictor of child development.
Response: Yes, maternal IQ might affect child development. Unfortunately, that data are not available. However, if we use the mothers' education levels as the indicator, we found no difference among the case and control groups (Table 1).

Comment 2:
Selection and recall bias: the authors state that "There may also have been a problem with recall bias during the interviews where participants may not have recalled or did not know the name of the pesticides they'd used in the past. However, if this did occur, it would have happened equally between groups. It was also very likely that study participants were exposed to pesticides in the environment, and this information bias could lower the strength of the reported association." but this is simply untrue. Mothers with children with SDD are far more likely to either recall precisely the chemicals that were used or to believe that the chemicals they used were the cause of SDD. Thus, I think this study is highly susceptible to recall bias. Similar issues with selection bias in that those that have more pronounced SDD were likely to volunteer or than those we less pronounced SDD or controls and the controls may have other uncontrolled differences.

Response:
Concerning the problem of mothers with children with SDD are more likely to have a better recall of chemical use and be more willing to participate in the study. Yes, we agree that the problem will seriously affect the study result. However, that problem will occur only when the mothers are known, or aware that pesticide is a potential cause of SDD, which might be the case in Thailand. In the country, currently, data on toxic effects especially chronic consequences, such as developmental delay are limited and not publicized. One aim of this study was to explore the association and bring greater public attention to the potential effects of pesticide use in SDD cases.
In addition, from our data, we found only a slightly higher participation rate among the case than that of the control (86.67% vs. 80.98%).
This information has been added to the result section.

Comment 3:
Was information on crop provided by the mothers? This may have allowed the authors the opportunity to validate some of their questions.

Response:
It is a good point. Unfortunately, we did not have that information. From what we know, most of the people in this area are rice farmers.

Comment 4:
The biggest limitation in this study is the retrospective exposure assessment via questionnaire. Ever recall using pesticides is likely quite imprecise and mothers may not wish to divulge that they were exposed to pesticides during pregnancy so as not to be "blamed" for their child's SDD. Exposure misclassification is not mentioned in the limitations yet all of the studies in the discussion including the one central Thai study used more objective measures of exposure (i.e., biological levels). When comparing the studies, the authors do not mention this. Similarly, the authors cannot derive any exposure magnitudes from this study which could obscure any underlying associations.

Response:
Yes, we agree that using exposure assessment via questionnaire might have some limitations. However, it may be the best tool for long-term exposure assessment, especially for a large survey study. Measurement of a biomarker in blood or urine is costly and represents only short-term exposure. This questionnaire technique was used in a wellrecognized Agricultural Health Study in the United States [1] and many other previous studies.
A few changes have been made and this information added to the discussion.
The problem of fearing to be blamed for using pesticides during pregnancy may not occur in this study because, as we mentioned before, the mothers were not aware of pesticides as a potential cause of SDD. Therefore, we believed that exposure and misclassification might not be a big problem, and, if occurred, it would be equally distributed among the case and control groups.
Yes, it is a good idea to explore the effect of exposure magnitudes when data are available.
In this study, we try to collect data on the duration and intensity of pesticide use. However, since only a small number of mothers reported using pesticides during pregnancy, the data was too small for further analysis.
Exposure misclassification has been added to the discussion.

Comment 5:
Because the authors findings differ so much from other studies that prospectively evaluated exposure and disease, they should elaborate more on why this is likely. For example, most prospective studies have found that prenatal exposures mostly drove any developmental delays while early childhood exposures had much less of an impact which is directly opposite of what the authors found. In all likelihood, the differences are likely because of exposure misclassification because the other studies had better exposure assessments1,2,3 ,4,5.

Response:
Thank you for raising this important issue and sorry for causing any misunderstanding.
First of all, we have to correct the point that our results are different from others. This is not true. Our results were consistent with the literature. In this study, we found that many pesticides especially chlorpyrifos was strongly associated with SDD. Concerning the window of periods of exposure, either prenatal or postnatal exposure to chlorpyrifos increased the risk of SDD. However, due to, maybe the small sample size, only the postnatal group showed a higher and more significant association. Currently, only a few studies reported data on both prenatal and postnatal exposure, further discussion is thus limited. By concept, both periods of exposure could affect SDD but greater effects are expected for prenatal exposure.
More information has been added to the discussion.

Comment 6:
It would be useful for the authors to let us know how their populations compare in demographic characteristics as the population in Northern Thailand or all of Thailand and how this may differ from other populations in the study.

Response:
Sorry, we don't have the data.
However, based on previous studies, it was found that factors affecting children with SDD were the parents' age, education level, occupation, income, and alcohol consumption [2]. However, in this study, the case was matched with controls by age and gender. In addition, those risk factors were included in the regression model to minimize the effect on the association.

Ru-Lan Hsieh
Department of Physical Medicine and Rehabilitation, Shin Kong Wu Ho-Su Memorial Hospital, Taipei Medical University, Taipei, Taiwan I have some comments for this revised paper, as listed below: It is surprising that the ages of mother ranged from 13-42 in controls and 13-46 in cases. Please provide the numbers of ages less than 18 in both groups, respectively, in Table 1. 1.
Please provide the legal age for marriage in Thailand, and re-analyze the data by age less than 18 and >=18, in addition to the continuous data as analyzed. 2.
The only significant factor to suspected developmental delay is Chlorpyrifos exposure in the present study. After the correction of crude OR in Table 3, the crude OR of ever Chlorpyrifos exposure is 1.11, prenatal exposure is 1.03, and postnatal exposure is 1.88. After adjustment, the OR of ever Chlorpyrifos exposure is 3.71 (p<0.05), prenatal exposure is 2.97, and postnatal exposure is 5.92 (p<0.05). The ORs after adjustment were much higher than the crude OR. Please provide the detailed regression model and correlation coefficient of each variable. 3.
The authors did not adjust the factor of pesticide environment exposure for OR, why? 4.
How could the authors get the contents of the pesticide as detailed as 14 kinds (such as Glyphosate, Paraquat, Butachlor, methyl parathion, Savin, Chlorpyrifos, etc.) just by questionnaires?

5.
In the discussion section, the authors mentioned that "This is consistent with the present study which also found an elevated risk of SDD among those exposed to mancozeb, with odds ratio of 1.87 for prenatal exposure, and OR of 3.97 for postnatal exposure (Table 3)". However, the OR did not reach statistically significant (p<0.05) in the present study. Therefore, the authors should not give this comment.

If applicable, is the statistical analysis and its interpretation appropriate? Partly
Are all the source data underlying the results available to ensure full reproducibility? Partly Are the conclusions drawn adequately supported by the results?

Partly
Competing Interests: No competing interests were disclosed.
Reviewer Expertise: Rehabilitation medicine; pediatric rehabilitation; developmental delay I confirm that I have read this submission and believe that I have an appropriate level of expertise to state that I do not consider it to be of an acceptable scientific standard, for reasons outlined above.

Author Response 05 May 2021
Chudchawal Juntarawijit, Naresuan University, Phitsanulok, Thailand I have some comments for this revised paper, as listed below: 1. It is surprising that the ages of mother ranged from 13-42 in controls and 13-46 in cases. Please provide the numbers of ages less than 18 in both groups, respectively, in Table 1.

Response:
First, we would like to thank you for your time and efforts provide valuable comments and suggestions.
As suggested, the numbers of participants aged less than 18 in each group have been added to Table 1. It can be noticed that the proportion of participants with <18 years in each group was similar to and there was no significant difference between the age group and developmental delay (p = 0.305).
2. Please provide the legal age for marriage in Thailand, and re-analyze the data by age less than 18 and >=18, in addition to the continuous data as analyzed.

Response:
In Thailand, the law restricts marriage under the age of 17. In this study, we started with selected children with developmental delay and their controls. The information of the mother and pesticide exposure were collected later, without restriction on either the marital status or age of the mother.
Further analysis using age as a categorical variable (<18, and 18 or more) was conducted as a suggestion, and the results, OR (adjusted2 in table 1-1). However, not much difference was observed as compared to the previous result when age was treated as a continuous variable (Table 3). It was also found that there was no literature to support a higher risk of developmental delay and among mothers below 18 years of age. Under the circumstances, therefore, it is better to treat age as a continuous variable. Table 3, the crude OR of ever Chlorpyrifos exposure is 1.11, prenatal exposure is 1.03, and postnatal exposure is 1.88. After adjustment, the OR of ever Chlorpyrifos exposure is 3.71 (p<0.05), prenatal exposure is 2.97, and postnatal exposure is 5.92 (p<0.05). The ORs after adjustment were much higher than the crude OR. Please provide a detailed regression model and correlation coefficient of each variable.

Response:
The difference between crude and adjusted OR might be explained by the fact that there are several factors that can affect child development. The results support that control of confounding factors is important. In this study, the adjusted variables were: mother's age when pregnant (continuous), education (no school, primary school, secondary school, college degree), occupation (farmer, own business, civil servant, employee [formal], employee [general work], housewife, retired, unemployed), income (<5000 baht, 5000-9999, 10000-14999, 15000 or more), chronic disease (yes, no), alcohol consumption (yes, no), gestation (<37 weeks, 37 or more weeks), birth order (1, 2, 3 or more), delivery method (vaginal delivery, cesarean section, assisted delivery), baby weight (<2500 grams, 2500 grams or more), and breastfeeding (yes, no). This information had been already presented in the Method section and in the table. In Table 3-1, the output of regression analysis of chlorpyrifos exposure and the correlation coefficient (B) of each variable was presented.

The authors did not adjust the factor of pesticide environment exposure for OR, why?
Response: This is a good point and thanks to you for the question.
Yes, we did not adjust for pesticide environment exposure for several reasons. One reason is that the model has already contained many adjusted variables whereas the sample size was rather small with only 47 control and 46 cases reported using pesticides. It was also found that all of the environmental exposure variables were not significantly associated with developmental delay (Table 1). In addition, further analysis shows that these variables were strongly associated with pesticide use (data was not presented). For example, chlorpyrifos use was strongly associated with having a family member working as a farmer (p<0.001), year of residency in the area (p=0.010), frequency of farm entry (<0.001), and keeping pesticides in a house (p=0.009).
5. How could the authors get the contents of the pesticide as detailed as 14 kinds (such as Glyphosate, Paraquat, Butachlor, methyl parathion, Savin, Chlorpyrifos, etc.) just by questionnaires?

Response:
Yes, the data on the 14 pesticides were collected using a face-to-face interview questionnaire. During the interview, village health volunteers asked each participant whether or not he/she ever used each individual pesticide, such as glyphosate, paraquat, butachlor, etc. This data collection method has been accepted and widely used in epidemiological studies. The method was suitable for a large study and studies that looked for the effects of long-term exposure because the blood or urine analyses are expensive and represent short-term exposure only.
6. In the discussion section, the authors mentioned that "This is consistent with the present study which also found an elevated risk of SDD among those exposed to mancozeb, with odds ratio of 1.87 for prenatal exposure, and OR of 3.97 for postnatal exposure (Table 3)". However, the OR did not reach statistically significant (p<0.05) in the present study. Therefore, the authors should not give this comment.

Response:
Yes, it was true that the OR was not significant. However, when considered an association or effect, it should not depend solely on p-value and use it as a magic number. The problem of p-value and its misuse was discussed quite frequently in the literature. For example, Wasserstein, Schirm, and Lazar (2019) said that "Don't believe that an association or effect is absent just because it was not statistically significant." In the paper of Bonner et al. (2017) published in Environmental Health Perspective, many of the associations between pesticides and lung cancer were presented when the OR or HR did not reach statistically significant, p <0.05 or confident interval included 1. For example, "the association between pendimethalin use and lung cancer (HR = 1.50; 95% CI: 0.98, 2.31)".

If applicable, is the statistical analysis and its interpretation appropriate? Partly
Are all the source data underlying the results available to ensure full reproducibility? Partly

Are the conclusions drawn adequately supported by the results? Partly
Chudchawal Juntarawijit, Naresuan University, Phitsanulok, Thailand

Reviewer II
This study aimed to evaluate the association between pesticide use by mothers during pregnancy and preschool children development using a case-control design in Thailand. They concluded that chlorpyrifos exposure during pregnancy was positively associated with developmental delay in children less than 5 years. I have some comments as listed below: 1. The cases of children included in the present study were "suspected developmental delay" rather than "confirmed developmental delay". Therefore, it would severely affect the results.
We agree with the reviewer that using confirm developmental delay may be the best outcome variable for studying conclusion. However, a number of children with developmental delay in the study area was so small, we have to rely on Suspected Developmental Delay (SDD). SDD is an important outcome variable and it has been used widely either in public health surveys for early identification of the prevalence rate, or to study risk factors. For example, a large study in 8 counties of rural China by Yang et al. (2019) 1 published in BMC Pediatrics, use SDD variables to evaluate the effects of care quality and development of children aged 1-59 months. Another study by Valla and team (2015) 2 also used SDD to study prevalence rates of developmental delay in Norwegian infants.

2.
As the authors pointed out, there were only 47 (10.4%) of case mothers and 46 (11.4%) of control mothers reported ever using any pesticide during pregnancy. The case numbers were too small for comparison.
Yes, we admitted that sample size is the study's limitations. The number of mothers ever using pesticides is actually the exposure of interest. In this study, we collected data from about 800 participants, the minimum sample size required for this type of study, under the following assumption: 95% confident interval, power of detection = 80%, case to control = 1:1, odds ratio = 1.5, and control with exposure = 40% (as presented in the methods section). Unfortunately, only about ten percent of the mother had experience using pesticides. The problem was beyond our control, and with some other constraints, we have to report the result as such.
We did not completely agree that the sample size was 'too small'. The issue is rather subjective, and it should depend mainly on the purpose of the study, and the statistics used for the analysis. The problem of using a small sample size is the lack of power of detection and precision. As seen in the study results, OR and other statistical parameters will not be significant. A small sample size will not completely destroy the usefulness of the study if it was analyzed with proper statistics and clearly presented.
According to the following article (DOI: 10.22004/ag.econ.103771), sample size should not be a main concern for logistic regression. Thus, we believed the result is good enough to be presented to research community.
As said by P. Mean. "A small sample size does not mean that your results are "wrong". It means that the data is consistent with a wide range of possible hypotheses." ( http://www.pmean.com/11/WideInterval.html) 3. There were only 11 (2.7%) case mothers and 13 (2.9%) of control mothers had ever exposed to chlorpyrifos. Therefore, using these very small numbers of exposure to chlorpyrifos to evaluate the risk of developmental delay of children is not adequate at all. The results would severely mislead the readers.
As mentioned before, the minimum sample size depends on the kind of statistic used for data analysis. If the sample size is too small, the OR obtained was not significant, as seen in many individual pesticides. This confirmed that the study results will not mislead the readers. We do the best to present study results. Data was analyzed with appropriate statistic, and the results were widely discussed. Therefore, readers can justify by themselves the reliability of the results. . The age ranges of mothers were between 13-42 and 13-46 in the control group and case group, respectively. Please re-analyze the mother's age by below 18 (or 20) vs. above 18 (or 20).
Thank you for suggestions, it is a good point. At first, we considered the best age to have a healthy baby is between 25 and 35 years of age, and thus using 25 as a cut point. However, actually, there is no scientific data to support the idea. So, we agree and decide to recategorize the age group to be <20, 20-25, 26-30, 31-35, and ≥36 years, as suggested.
5. The only significant variable to developmental delay was the exposure to chlorpyrifos, and other variables were not statistically significantly associated with developmental delay. The authors calculated the crude odds ratios of chlorpyrifos ever exposure was 2.88, prenatal exposure was 2.34, and postnatal exposure was 4.18 as shown in Table 3. However, I found that it should be 1.10 in ever exposure, 1.02 in prenatal exposure, and 1.87 in postnatal exposure. Please recheck all variables' odds ratios carefully.
Thank you so much for your effort to identify the problem. All the data was checked and the error was found only with the crude OR. Data in Table 2 and Table 3 has been revised. Sorry for the mistake.
assessment (description) is in question, only the questionnaire was used and there was obvious recall bias. How we can ensure the reality of pesticide exposure history? The authors should add more information on this point.
Besides, several pesticide exposures were reported to have association with suspected developmental delay, what about the biological mechanism? Please add more literature on animal studies to support your results. Currently, only the similar studies were mentioned.
Furthermore, the following minor comments should also be considered: The description that "Three records were not included for analysis because important information such as gender and age, was missing" in the section of results (Page 4) can't be understood, since you have data of all cases firstly, then select the reference.
amounts, use way, etc.) in these areas during the period when the mother were in pregnancy? It is important to use such data to confirm the answer of these mothers, specifically the use stage (pre-N, or post-N).
We agree that the data will be useful. Unfortunately, there was no such data in the area, especially data of individual pesticides. As we mentioned before, using a questionnaire may be the best and the only way to collect data on long-term historical exposure of pesticides.