Association between pesticide exposure and paraoxonase-1 (PON1) polymorphisms, and neurobehavioural outcomes in children: a systematic review

Articles considered for the review were published between the years 2000 and 2018. Articles were retrieved using online search engines and library sources, including the Institute for Scientific Information (ISI) Web of Knowledge, Cochrane library, Google, Google Scholar, PubMed Search and Science Direct. The key terms used to generate the search were “pesticide exposure” and “prenatal exposure”, “organophosphates/organophosphorus” and “single nucleotide polymorphisms”, “paraoxonase-1 (PON1)”, and “neurodevelopment/neurobehavioural outcome in child/infant”. Individual terms and a combination of terms were used to search for articles. Unpublished studies and the grey literature were not reviewed for this search. Reference lists of retrieved articles were also reviewed, and relevant articles were retrieved.

Exclusion criteria were (a) articles published prior to the year 2000, (b) articles in languages other than English were excluded, (c) articles including pesticides other than organophosphates were excluded, (d) if the evidence of pesticide exposure was unclear or not documented the article was excluded, (e) articles reporting on SNPs in genes other than the PON1 gene were excluded and (f) non-neurobehavioural health outcomes were excluded from the search (Table 1).

The literature search strategy was developed by Nkosinathi Banhela (NB) who is the first author on this paper and was approved by his PhD supervisor, Dr Saloshni Naidoo (SN) a co-author on this paper. Articles were screened for inclusion eligibility in this study by NB and confirmed by PN. Two reviewers (NB and PN) reviewed the title, abstracts and full-text for eligibility of the studies to be included in the current study, and a third reviewer checked for discrepancies (SN). The online search identified 179 relevant citations. The titles and abstracts of 76 articles were reviewed in detail; 103 articles were reviewed not to be in line with the inclusion criteria. After a thorough review of the titles and abstracts of articles, the full-texts of 49 articles were retrieved and screened; 27 articles were not in congruency with the inclusion criteria. Six articles met the inclusion criteria. Figure 1 summarises the online search using the PRISMA flow diagram. The reference manager tool that was used to record and manage the referencing and the reference list is “Endnote X9” referencing managing software (Table 2).

Data from the selected articles were extracted into an Excel spreadsheet. The Excel spreadsheet was structured to visualise alignment of selected articles with the inclusion and exclusion criteria. NB performed the initial synthesis of collected data; PN verified and checked the reported findings on data. Data were extracted relating to the study title, authors, publication year, sample size, study design, biological sample analysed, neurobehavioural tests used, age of children assessed and outcomes of neurobehavioural tests (see Table 3).

Data from articles selected to be included were tabulated to represent study sample size, study design, screening tool used to measure pesticide of interest, genotype of interest, neurodevelopmental tool used and age it was administered. The above-mentioned parameter was then compared among the articles selected for inclusion in the study. Further, the PON1 genotype was associated with the observed/screened neurobehavioural outcome and the statistical score used for the associations (see Table 4). Neurobehavioural outcomes explored in each selected article were characterised to have either cognitive, behavioural, sensory, motor and morphological effect. A meta-analysis for the current available data could not be performed as there were no randomised controlled trials found in the literature. While the health outcomes under study were in the same category of neurodevelopmental outcomes, different measuring instruments were used by the studies. The neurobehavioural tests in the different papers rendered results which were either behavioural, cognitive, sensory, motor or morphological, making it difficult to compare for a meta-analysis.

The “Quality Assessment Tool for Quantitative Studies” developed by the Effective Public Health Practice Project (EPHPP) was used to appraise the quality of articles for inclusion in the review [30]. These guidelines assess the quality of studies based on selection bias, study design, confounders, blinding, data collection tools, withdrawal and drop-outs, intervention integrity and analysis. The first author (NB) performed the quality appraisal for articles that were in-line with our current study; then, the reviewer (PN) performed the quality appraisal according to his judgement for the selected articles. Reviewers then cross-checked if the appraisal had any differences which were settled by the third reviewer (SN) (as she is the author that is most experienced in this study). Studies which showed poor quality (< 4 score) were excluded from the review. Articles with a moderate to a strong rating were included in the study (see Tables 2, 5, and 6).

In four of the articles included in this review, the studies followed a cohort study design [5, 21, 26, 29], whilst the remainder two articles followed a case-control study design [21, 28]. Two articles emanated from one study [5, 21]. Four of the papers reported on studies in the USA [12, 18, 21, 22]; one study involved Italian and Canadian participants [27] and one study was based in Romania [28]. Quality assessment of selected articles indicated that most of the articles selected for the study were of a good quality rating, except the article by Pasca et al. which did not report on pesticide exposure.

Four articles reported on mother and child pairs [5, 21, 26, 29], whilst one study reported on children [28]. The age of children assessed ranged from 1 to 9 years in the selected articles. One article assessed children at three stages (one, two and 6 to 9 years) [29], whilst another article assessed children at two stages (5 and 7 years) [21]. One article assessed children at only one age point, at 2 years [25] and one article at 8 years [26]. The age of children in one article was not reported [28].

In three articles [5, 21, 29], urinary samples were taken to assess prenatal pesticide exposure. In one article [27], urinary peptides were measured. In all the articles, blood samples were taken for PON1 SNP genotyping and enzyme assessment. Four articles [5, 21, 26, 29] included in this review explored pesticide load using gas chromatography tandem mass spectrometry (GC/MS). One article [27] used high performance liquid chromatography (HPLC) [25] whilst another used an enzymatic activity assay [28] to quantify pesticide exposure.

Four articles reported on the PON1 (108T) SNP [5, 21, 26, 27], whilst two articles reported on the PON1 (L55M) SNP [27, 28]. All six articles reported on the PON1 (Q192R) SNP. Other SNPs which have been explored by researchers were not included because of their associated health outcome being outside the scope of this study. It was deduced that various SNPs exhibit significant inter-ethnic differences, meaning their distribution differs among different ethnic and cultural groups [22].

Children assessed for neurobehavioural health outcomes were between the age(s) of 1 to 9 years old. Tools used to assess neurobehavioural health outcomes in the included studies were the Bayley Scale of Infant Development (BSID) (two articles [5, 29]) which measures the Mental Development Index (MDI) and Psychomotor Development Index (PDI), the Conners’ Kiddie Continuous Performance Test (K-CPT) (one article [21]), the Wechsler Intelligence Scale (WISC) for Children (one article [21]) and the Behaviour Assessment System for Children-2 (BASC2) (one article [26]). Two articles included in the review [27, 28] reported on autism spectrum disorder (ASD) and used a physician diagnosis of autism using the standard approved diagnostic method.

Three articles reported a significant association between organophosphate exposure, PON1108T and Q192R polymorphisms, and neurobehaviour health outcomes [12, 18, 22]. No significant associations were observed in the remaining three articles [23‐25]. Eskenazi et al. quantified maternal urine dialkyphosphate (DAP) levels, genotyped the PON1 gene for PON1 192 and PON1 108 polymorphism and measured the enzymatic activity of paraoxonase (POase) and arylesterase (ARYase). The PON1 108T allele in children was associated with poorer MDI scores and high DAP levels [5]. However, other PON1 polymorphic variants and enzymes showed no significant association. The association of the PON1 108TT allele with poorer/low neurofunction scores was further reported in a follow-up study of the same cohort [21]. In this paper [22], maternal PON1 108T allele carriers showed a weakly modified relationship between DAP and neurobehaviour. Engel et al. found the PON 1 Q192R QR/RR variant genotype to be associated with poorer perceptual reasoning and increasing dialkyl- and dimethylphosphate (DAP and DMP) concentrations in urine. The adverse neurobehavioural health outcome in children was observed from 12 months of age and continued up to 9 years of age [29]. The low enzymatic activity was associated with the PON1 Q192R allele and with exposure to DAP and DMP [21]. ARYase and POase levels and activity were also reported to be associated with the neurodevelopment hindrance, leading to autism spectrum disorders in Romanian children [28]. However, the above-mentioned study did not find any significant association in PON1 allelic distribution [28]. The PON1 R192 gene variant was implicated in the pathogenesis of autism among North American (Caucasian), but not Italian, organophosphate users [27].