MOF, Alm.del - 2020-21 - Bilag 709: Orientering om nyt om befolkningsundersøgelser om mulige sundhedseffekter ved udsættelse for sprøjtemidler, fra miljøministeren

Miljø- og Fødevareudvalget 2020-21
MOF Alm.del Bilag 709
Offentligt

NOTAT

Pesticider og biocider

J.nr. 2020 - 11609

Ref. perrj

Den 14. april 2021

EFSA’s

vurdering

af mulig sammenhæng mellem ADHD-

symptomer og metabolitter af pyrethroider, baseret på Odense

Børne Kohorten

Problemstilling

En forskningsartikel, hvor der rapporteres om en mulig sammenhæng mellem ADHD-symptomer og

metabolitter af pyrethroider baseret på Odense Børne Kohorten blev offentliggjort i 2019

Miljø- og Fødevareministeriet har bedt om en yderligere kvalificering af forskningsartiklen om en

mulig sammenhæng mellem ADHD-symptomer og urinmetabolitter af pyrethroider og chlorpyrifos i

henhold til den fastlagte praksis for reguleringsmæssig brug af befolkningsundersøgelser om effekter

af sprøjtemidler, jf. https://www.ft.dk/samling/20191/almdel/MOF/bilag/386/index.htm , idet

artiklen falder i mellemgruppen, hvor der er et middelstærkt grundlag for reguleringsmæssige indgreb.

DTU’s

vurdering af artiklen er fremsendt til departementet den 9. januar 2020. Vurderingen viste, at

undersøgelsen var af ”lav/middel pålidelighed” og med ”nogen evidens for en årsagssammenhæng”.

Miljøstyrelsen har efterfølgende bedt forfatteren til artiklen om kommentarer til den faglige vurdering

fra DTU. Miljøstyren har modtaget forfatterens kommentarer og sendt dem til DTU med henblik på at

vurdere, om kommentarerne giver anledning til at ændre vurderingen af artiklen. DTU’s opdaterede

faglige vurdering af artiklen fra den 29. juni 2020, indeholdt

meget få tilretninger. DTU’s

opdaterede

vurdering af artiklen er fremsendt til departementet den 3. juli 2020.

Miljøstyrelsen har endvidere sendt artiklen til EFSA mhp. at vurdere studiet og informere om evt.

igangværende EU-arbejde vedrørende aktivstoffer, som danner de pågældende

nedbrydningsprodukter.. MST har endvidere anmodet om, at EFSA vurderede statistikken og

robustheden af fundene. EFSA har vurderet studiet på deres 108. PPR Panel møde den 19. november

2020, hvor Miljøstyrelsen og DTU også deltog. EFSAs fulde vurdering samt svar på uddybende

spørgsmål fra Miljøstyrelsen fra 14. december 2020 og 22 januar 2021 følger herunder.

Det fremgår, at PPR panelet vurderer, at de statistiske metoder, som er anvendt er hensigtsmæssige.

PPR Panelet vurderer endvidere, at der er en mulig sammenhæng mellem prænatal eksponering for

pyrethroider og forældrevurderet ADHD-problemscore, men at der ikke er vist årsagssammenhæng.

Samlet set er det EFSA’s vurdering, at

de eksisterende epidemiologiske beviser for sammenhængen

mellem urin-niveauer af 3-PBA (og trans-DCCA) og ADHD er ret begrænsede og, da modstridende

Dalsager L, Fage-Larsen B, Bilenberg N, Jensen TK, Nielsen F, Kyhl HB, Grandjean P, Andersen HR.

Maternal urinary concentrations of pyrethroid and chlorpyrifos metabolites and attention deficit

hyperactivity disorder (ADHD) symptoms in 2-4-year-old children from the Odense Child Cohort.

Environ Res. 2019; 176: 108533. doi: 10.1016/j.envres.2019.108533

navn

•

adr1

•

adr2

•

adr3

•

postnr by

Tlf. Nr

•

Fax nr

•

CVR nr

•

EAN nr

•

web

MOF, Alm.del - 2020-21 - Bilag 709: Orientering om nyt om befolkningsundersøgelser om mulige sundhedseffekter ved udsættelse for sprøjtemidler, fra miljøministeren

resultater også er set i litteraturen for 3-PBA (hovedmetabolitten), at der er behov for flere

undersøgelser med optimalt design og udført i forskellige populationer for at underbygge fundene.

EFSA vurderer, at resultaterne for Trans-DCCA (en mindre metabolit, fundet i 11%), både kan

repræsentere en ægte effekt eller alternativt en tilfældig effekt på grund af det relativt lille antal børn,

der havde en score for ADHD symptomer større end 90-percentilen, og hvis mødre samtidig havde

trans-DCCA niveauer over detektionsgrænsen. Sandsynligheden for at et statistisk signifikant fund

afspejler en reel effekt påvirkes negativt af lav statistisk styrke i et studie, som det er tilfældet i dette

studie.

EFSA’s

Vurdering af Dalsager et al, 2019 artiklen fra EFSA PPR Panelet, modtaget pr

mail den 9. december 2020:

”

Study under evaluation:

Dalsager L, Fage-Larsen B, Bilenberg N, Jensen TK, Nielsen F, Kyhl HB, Grandjean P, Andersen HR.

Maternal urinary concentrations of pyrethroid and chlorpyrifos metabolites and attention deficit

hyperactivity disorder (ADHD) symptoms in 2-4-year-old children from the Odense Child Cohort.

Environ Res. 2019; 176: 108533. doi: 10.1016/j.envres.2019.108533

Question raised by Pernille Rosenskjold Jacobsen (Ministry of Environment and Food

of Denmark)

We kindly ask EFSA to consider the paper (Dalsager et al, 2019) and evaluate the associations

(reliability and causality) observed for pyrethroids. Especially relating to the robustness of the

association observed between ADHD symptoms and the different metabolites considering the size of

the increases (odds ratio and relative change ratio) observed, the quality of the methods and the

statistical analysis used in the study.

Summary of the study

This study assessed the association between prenatal exposure to pyrethroids and chlorpyrifos and

traits of ADHD in 2-4-year-old children from the Odense Child Cohort (Denmark). The main results

showed that every doubling in maternal 3-PBA concentration was associated with a 3% higher than

expected ADHD score (adjusted ratio: 1.03; 95% CI: 1.00–1.07). Also, a doubling in maternal 3-PBA

concentration was associated with 13% higher odds of having an ADHD score ≥ the 90

percentile

(adjusted OR: 1.13; 95% CI 1.01–1.25).

The odds of an ADHD score ≥ the 90

percentile was 1.76-fold

higher in children whose mothers had trans-DCCA concentrations above LOD compared to children

whose mothers had concentrations below the LOD (adjusted OR: 1.76; 95% CI: 1.08–2.86).

Concurrent concentrations of 3-PBA and the chlorpyrifos metabolite TCPY above their medians were

associated with higher ADHD score (adjusted relative change ratio: 1.20; 95% CI 1.04–1.38) and

higher odds of scoring ≥ the 90th percentile (adjusted OR: 1.98; 95% CI 1.26–3.11).

The study

concluded that prenatal exposure to pyrethroids was associated with ADHD related traits at 2–4 years

of age.

Assessment of the study

Overall, the study of Dalsager et al. (2019) only supports a

possible

association between prenatal

exposure to pyrethroid insecticides metabolites and the prevalence of ADHD traits in childhood. The

strengths, limitation and risk of bias of the study are briefly addressed below.

1. Main

strengths

of the study:

MOF, Alm.del - 2020-21 - Bilag 709: Orientering om nyt om befolkningsundersøgelser om mulige sundhedseffekter ved udsættelse for sprøjtemidler, fra miljøministeren



The prospective follow-up design (birth cohort) and the large sample size (n=948) are the

main strengths of this study. The advantage of cohort studies is that exposure is assessed

before the occurrence of the outcome measured. Accordingly, the reliability of this study is

higher than that of a few other studies addressing the association between urinary levels of 3-

PBA and ADHD traits, such as Lee et al. (2020)

, Wagner-Schuman et al. (2015)

and

Richardson et al. (2015)

, which used a cross-sectional design.

The use of biomonitoring for metabolites of the exposure of interest is generally considered to

be the most relevant and reliable way to assess exposure in epidemiology studies. In this study,

a number of specific and non-specific pyrethroid metabolites were measured, albeit with

varying analytical sensitivity.

2. The major

limitations

of this study are the following:



Possible misclassification of the exposure over pregnancy as only a single urinary sample was

obtained at gestational week (GW) 28. Although the study claims that the exposure level is

assumed to be more stable in populations continuously exposed to low concentrations of

insecticides from residues in the diet, this assumption has not been demonstrated.

Potential for misclassification of ADHD scores as these were parent-reported and the severity

of symptoms may be systematically under- or overestimated.

The size of the effect when analysing for the expected relative change (ratio) in ADHD score

was rather low (3% and 13%) when 3-PBA was expressed as a continuous variable.

Furthermore, the results were significant only for girls (relative change ratio 1.06, 95% CI

1.00–1.12) but not for boys (1.02; 95% CI 0.98–1.06). Effect sizes were similar but slightly

larger for the OR for

scoring ≥ the

90th percentile on the ADHD scale. However, only 18% of

assessed children fell into this category.



3. The internal validity or

risk of bias

(RoB) of the study was appraised using a customised version of

the OHAT/NTP RoB assessment tool (Handbook for Conducting a Literature-Based Health

Assessment Using OHAT Approach for Systematic Review and Evidence Integration, 2015

). This tool

was developed to provide an approach to the evaluation of RoB in the context of hazard identification

for human risk assessment of chemicals. The following seven questions were assessed and rated as for

RoB



Q1: Did the selection of study participants result in appropriate comparison

groups?

Rated as definitively low RoB. The study participants (mother-child pairs) were derived from a

prospective birth cohort study, pregnant women were recruited from the same eligible population,

using the same method and inclusion and exclusion criteria

DOI: 10.1016/j.envres.2020.109739

DOI: 10.1186/s12940-015-0030-y

DOI: 10.1096/fj.14-260901

Handbook for Conducting a Literature-Based Health Assessment Using OHAT Approach for Systematic Review

and Evidence Integration. Office of Health Assessment and Translation (OHAT) Division of the National

Toxicology Program National Institute of Environmental Health Sciences.

https://ntp.niehs.nih.gov/ntp/ohat/pubs/handbookjan2015_508.pdf

MOF, Alm.del - 2020-21 - Bilag 709: Orientering om nyt om befolkningsundersøgelser om mulige sundhedseffekter ved udsættelse for sprøjtemidler, fra miljøministeren



Q2: Did the study design or analysis account for important confounding and

modifying variables?

Rated as probably low RoB. Factors (covariates) that may influence either gestational maternal

exposure to insecticides or ADHD symptom score in the children were included in a directed

acyclic graph DAG

. Negative regression models were adjusted for maternal educational level,

parental psychiatric diagnosis, child age and sex, 3-PBA and TCPY concentrations, and urinary

creatinine. However, potential co-exposure to developmental neurotoxicants (e.g.,

organophosphorus compounds, lead, mercury, among others) was not measured with the

exception of urinary TCPY (a relatively specific metabolite of chlorpyrifos).



Q3: Were outcome data complete without attrition or exclusion from analysis?

Rated as probably low RoB, particularly because of attrition. The CBCL: 1�½-5 questionnaire was

mailed to 2551 parents and a total of 1942 (76.6%) answered it. Only 1515 pregnant women

provided a urine sample during pregnancy (N=1515). Finally, the study population consisted of

948 participants. Potential selection bias was assessed and it was found that study participants

were generally older and more often non-smokers as compared to both groups of non-participants.

Also, participants providing a urine samples generally had a higher educational level as compared

to non-participants.



Q4: Can we be confident in the exposure characterisation?

Rated as probably high RoB. Exposure was assessed by monitoring for the common pyrethroid

metabolite (3-PBA) and other more specific metabolites in a single urine sample collected at GW

28. Analyses were performed by LC-MS/MS with the analytical method being based on a slightly

modified version of a previously reported method.

Only a single spot urine sample was collected in GW 28 and considered as a proxy for the pesticide

exposure level during pregnancy, which is not necessarily representative of pyrethroid exposure

over the entire pregnancy. Furthermore, a potential exposure after birth up to the age of

assessment at 27 months cannot be excluded.

Although the study determined 3-PBA, 4-F-3PBA, cis-DCCA, trans-DCCA, cis-DBCA in urine, only

3-PBA was detected in a high percentage of maternal samples (94.4%) and the remaining

metabolites were detected in a very small proportion of samples (trans-DCCA in 11.4% and the rest

in less than 5%). 3-PBA is a common metabolite of many synthetic pyrethroids, including

allethrin, cyhalothrin, cypermethrin, deltamethrin, fenpropathrin, fenvalerate, permethrin,

resmethrin, tralomethrin, and their isomers.

However, the use of 3-PBA as urinary biomarker to

assess exposure to these compounds is limited by the possibility that measured concentrations

may represent direct exposure to 3-PBA formed in food items from the degradation of these

pesticides in the environment or metabolisation by plants rather than exposure to the parent

compounds themselves. Hence, biomonitoring of these biomarkers in urine will lead to

overestimation and misclassification of exposure for risk assessments and epidemiologic studies,

particularly in scenarios of low level exposure such as occur in dietary and other non-occupational

settings (Chen et al., 2012

Directed acyclic graph (DAG) models are popular tools for describing causal relationships and for guiding

attempts to learn them from data. They appear to supply a means of extracting causal conclusions from

probabilistic conditional independence properties inferred from purely observational data.

doi:10.1289/ehp.0901275

doi: 10.1021/jf303116p

MOF, Alm.del - 2020-21 - Bilag 709: Orientering om nyt om befolkningsundersøgelser om mulige sundhedseffekter ved udsættelse for sprøjtemidler, fra miljøministeren



Q5: Can we be confident in the outcome assessment?

Rated as probably low RoB since the Child Behaviour Checklist for ages 1�½-5 years (CBCL: 1�½-5)

is considered a validated tool. It is a parent-report online questionnaire on which the child is rated

on various behavioural and emotional problems. However, from 100 questions of the CBCL that

addressed behavioural, emotional, and social problems, only 6 correspond to an ADHD problem

scale. Furthermore, CBCL can be used to screen for ADHD disorders in preschool children

population, but it is not a specific diagnostic tool for ADHD. Finally, the questionnaire was not

administered by experienced neuropsychologists.



Q6: Were all measured outcomes reported?

Rated as probably low RoB. The ADHD score was analyzed as ordinal data from CBCL:1�½-5

results, and also as a dichotomized variable (scoring ≥ 90th percentile on the CBCL:1�½-5)

using

appropriated regression analysis. Besides, 94 questions of the CBCL questionnaire were not

reported as they were not directly related to ADHD.



Q7: Were there other potential threats to internal validity? Were the statistical

methods appropriate?

Rated as definitely low RoB. The statistical methods used were appropriate. Bivariate and

multivariate (binomial and logistic regression analyses) were used. Crude and adjusted sizes of

effects were obtained.

When the algorithms of OHAT/NTP were applied to the aforementioned ratings, the study was

categorised as tier 3 (high RoB). The main driver for this categorization was Q4 (exposure

characterization).

Overall appraisal with regard to the questions raised by the Ministry of Environment

and Food of Denmark

Human observational studies are not designed to provide direct evidence of causality. Instead, the

observed associations must be critically examined with regard to the likelihood of a cause-effect

relationship. There are several aspects of this study which make it difficult to assess the observed

associations with regard to causality. First and foremost, a spot urine sample taken at only a single

time-point during pregnancy provides very little information about the overall exposure to pyrethroids

during pregnancy and none about exposure during the first two years of life. Secondly, although

confounders were adjusted for, relevant co-exposures to other potential developmental neurotoxins

were not assessed with the exception of chlorpyrifos. Thirdly, the CBCL:1�½-5 instrument is parent-

administered and contains only a small number of questions related to behaviour possibly associated

with ADHD. Fourthly, the observed relative change (ratio) in the primary model (negative binomial

regression) was very small (only 3% higher than expected). Although the auxiliary model (logistic

regression) showed similar and somewhat stronger associations when applied to those children

scoring

≥ the 90th percentile on the ADHD scale, this only applied to a relatively small subset of

children. It must therefore be concluded that an association between prenatal exposure to pyrethroids

and parent-assessed ADHD problem score is possible but no causality can be determined.

Overall, the available epidemiological evidence in humans on the association between urinary levels of

3-PBA and ADHD is rather limited (Dalsager et al., 2019; Lee et al., 2020; Wagner-Schuman et al.,

2015; and Richardson et al., 2015). Thereby, more studies with optimal designs and conducted in

MOF, Alm.del - 2020-21 - Bilag 709: Orientering om nyt om befolkningsundersøgelser om mulige sundhedseffekter ved udsættelse for sprøjtemidler, fra miljøministeren

different populations are needed to provide consistency to the available findings, as contradictory

results are also available in the literature (Quirós-Alcalá et al., 2014

).”

Miljøstyrelsen sendte EFSA yderligere uddybende spørgsmål den 14. december 2020.

Miljøstyrelsens spørgsmål efterfølges af EFSAs svar fra 22. januar 2021, markeret med

rød tekst.

Thank you for this very thorough assessment it is indeed very helpful. We just have a couple of

questions for clarification for Trans-DCCA, which was also discussed at the meeting.

Under Q4 it is stated “Although the study determined 3-PBA,

4-F-3PBA, cis-DCCA, trans-DCCA, cis-

DBCA in urine, only 3-PBA was detected in a high percentage of maternal samples (94.4%) and the

remaining metabolites were detected in a very small proportion of samples (trans-DCCA in 11.4% and

the rest in less than 5%).”

Does this mean that Trans-DCCA (as the other metabolites mentioned) is considered detected in so

few samples/few subjects that it is difficult to use the result?

Re: According to the study, trans-DCCA was detected in 11.4% of the study population (which

consisted of 948 mothers) whereas 3-PBA was detected in 94.4%. There are two possible reasons

underlying this large difference in the percentage of detection of both metabolites in urine samples:

•

The analytic sensitivity of the methodology was different for the metabolites studied. The LOD

was 0.03 ng/ml for 3-PBA and 0.4 ng/ml for trans-DCCA. Because pyretroids yielding trans-DCCA are

hydrolysed to one molecule of 3-PBA and another of trans-DCCA, the analytical sensitivity of trans-

DCCA was 13.64 times lower than that of 3-PBA on a molar basis.

•

While many type I and type II pyrethroids share 3-PBA as a common metabolite, trans-DCCA

only originates from a few pyrethroids (permethrin, cypermethrin and cyfluthrin). It is possible that

the study population had higher exposure to pyrethroids other than those yielding trans-DCCA.

Both metabolites have a similar apparent elimination half-life, 5.4 h for 3-PBA and 5.7 h for trans-

DCCA in volunteers following ingestion (doi:10.1093/annhyg/mev059).

Based on the lower number of mothers with trans-DCCA levels above the LOD, the study categorized

exposure as a dichotomous variable, i.e. above and below the LOD. In contrast to 3-PBA, the study

may be underpowered to detect a potential significant association of trans-DCCA with ADHD traits

due to a smaller sample size. In other words, the statistical power to detect significant associations was

much lower as compared to 3-PBA.

Otherwise trans-DCCA

is only mentioned in the initial “Summary of the study” where results from the

Dalsager et al. paper, are presented. The effects associated with detecting Trans-DCCA above LOD

does not seem to be considered in the overall appraisal although they were discussed during the

meeting?

Re: The appraisal paid more attention to 3-PBA because the study addressed the association of

pyrethroids (in general) with traits of ADHD. Based on the reasons mentioned above (i.e., a low

detection rate for trans-DCCA and potential for underpowered analysis) this metabolite was only

briefly considered for the study appraisal.

DOI: 10.1289/ehp.1308031

MOF, Alm.del - 2020-21 - Bilag 709: Orientering om nyt om befolkningsundersøgelser om mulige sundhedseffekter ved udsættelse for sprøjtemidler, fra miljøministeren

Miljøstyrelsen bad om yderlige uddybning af svaret den 22. januar 2020. EFSAs svar

kom den 1. Februar 2021 markeret med rød tekst:

The result of Trans-DCCA in Table 6 in the paper,

show the odds of an ADHD score ≥the 90th

percentile was 1.76-fold higher in children whose mothers had trans-DCCA concentrations above LOD

compared to children whose mothers had concentrations below the LOD (statistically significant and

also seen in boys and girls separately). Based on the answers in red is it correctly understood that,

because of the low power to detect significant associations for Trans-DCCA, the observed statistically

significant result in table 6 may actually not reflect a true effect?

There is a risk of over-interpreting single findings from observational studies, which should not be

taking out of context in order to rationalise them. No firm conclusions should be drawn from an

individual finding either, but the latter rather should be seen in the context of the whole study. That

finding may therefore be related to the inherent limitations of the study, which may have a greater

impact on individual parts thereof.

Having said that, the finding under consideration may represent a true effect observed in the study or,

alternatively, a random effect due to the relatively small number of children who had a CBCL- ADHD

score greater than the 90th percentile and whose mothers had trans-DCCA levels above the LOD. A

low sample size, a small effect size, or both may negatively affects the likelihood that a nominally

statistically significant finding of a study actually reflects a true effect as a result of low statistical

power (Button et al., 2013).