Introduction

Neuroblastoma (NB) is a malignant embryonal tumor of the neural crest cells that occurs almost exclusively in infancy and early childhood. Forty percent of NB cases are diagnosed before the age of one year and 85% before the age of five years [1].

The etiology of NB remains unknown. A genetic predisposition of NB has been suggested after the observation of rare familial cases (1% of NB cases) and a possible association with congenital malformations [2, 3]. Considerable progress has been made recently in the germ line and somatic genetic characterisation of patients and tumors [4]. However, genetic predisposition cannot fully explain its origin and other factors may be involved in NB development [5].

Considerable evidence suggests that NB may be initiated in utero during sympathoadrenal development [6]. The undifferentiated stem cells from the neural crest progenitors may persist in crest-derived tissues awaiting stimulation or reactivation in response to environmental or cellular cues [7]. Prenatal exposure to pesticides may be one such risk factor. A possible association between pesticides and NB was first suggested more than 30 years ago in relation to exposures to chlordane or heptachlor [8], which are active ingredients in many household and garden pesticides. An association between pesticide exposure and NB is plausible as more than 25 chemical compounds used in pesticides have been classed as potential human carcinogens [9, 10]. In addition, previous studies have showed that they pass through the placenta leading to fetal exposure [11, 12].

Literature about the subject is scarce. Only two previous studies [13, 14] have shown associations between NB and use of household pesticides during the preconception period, pregnancy, or in early childhood. Studies looking at parental occupational exposure to pesticides are heterogeneous in exposure definition and in findings. Some studies have suggested an increased risk of NB with maternal occupational exposure to pesticides during pregnancy [15] or at any time during the preconception-pregnancy period or childhood [14, 16], while a meta-analysis found no association with paternal occupational exposure to pesticides at any time [17]. A large cohort study [18] suggested an increased risk of NB in offspring of farm holders, while a registry-based study using geographic information [19] did not support an association with residence exposure to pesticides related to neighboring agricultural activities.

The aim of this study was to investigate whether the maternal use of household pesticides during pregnancy was associated with the risk of NB in her child. For these analyses, we pooled data from two large case–control studies, ESCALE and ESTELLE, which were conducted by the same investigators in France.

Materials and methods

Study population

The ESCALE and ESTELLE studies have been described in detail elsewhere [2]. Briefly, they were two nationwide population-based case–control studies, which were conducted with the support of the Société Française de lutte contre les Cancers de L’Enfant et l’Adolescent (SFCE).

The cases, which were directly identified from the French national registry of childhood cancer (RNCE), were children younger than 15 years old who lived in France at the time they were newly diagnosed with cancer. The ESCALE study included cases of NB, lymphoma, leukemia, and malignant brain tumor diagnosed in 2003–2004. The ESTELLE study included cases of NB, leukemia, lymphoma, childhood brain tumor, Wilms’ tumor, and hepatoblastoma diagnosed in 2010–2011.

Overall case participation rates were 81.2% for ESCALE and 92.2% for ESTELLE. Information on the MYCN amplification subtype was obtained subsequently from the RNCE.

The population controls were children free from cancer randomly selected from the French population using quota-sampling methods who were recruited by telephone during the same time periods. The participation rates were 71.2 and 85.5%, respectively.

In both studies, the cases and the controls were ineligible if their biological mother was unavailable, did not speak French, or had a serious psychosocial problem. In addition, 22 cases, 9 out of 248 cases under 18 months (3.6%) and 13 out of 252 cases aged 18 months or more (5.2%), were not eligible to maternal interview for ethical reasons because they had died or were in palliative care.

The present paper focuses on NB in children under six years old (91.3% of cases).

Data collection

Trained interviewers conducted standardized telephone interviews with the biological mothers of cases and controls, which lasted approximately 50 min. The interviews used similar scripts and were performed in the same conditions in the two studies. They focused on socioeconomic characteristics, prenatal and childhood environment exposures, familial and personal medical history.

In regards to pesticide exposures during pregnancy, the mother was asked if she used herbicides (“weed killers”), fungicides, or insecticides (and whether they were used indoors, for gardening or outdoors, or on pets). The mother was also asked if she was exposed to any type of pesticides in the workplace.

The ESTELLE study included additional questions about maternal pesticide use in the three months prior to conception and after birth, and whether there had been any professional pest control treatments of the home.

Data analysis

Study-specific odds ratios (ORs) and pooled ORs and 95% confidence intervals (95% CIs) were estimated by unconditional logistic regression (SAS version 9; SAS Institute Inc., Cary, NC, USA). All the models included the study matching factors: child’s age and gender, and, for the pooled analyses, the indicator of the study of origin.

The socioeconomic variables tested as potential confounders were: maternal level of education, size of the urban unit of residence, maternal age at child’s birth, birth order, and the type of housing during pregnancy. We also tested fetal growth, congenital malformations, and breastfeeding that were significantly associated with NB in previous analyses [2]. In the final model, only maternal age, birth order, size of the urban unit of residence, and the type of housing were retained.

Between-study heterogeneity was systematically tested by fitting an interaction term between the study and the exposure of interest.

We performed additional analyses stratified by age at diagnosis (<18 months/≥18 months), MYCN status (amplified/non-amplified), urban/rural status of the area of residence, and maternal level of education (less than baccalaureate/baccalaureate or higher).

Due to a small number of children with a congenital malformation, possible confounding was accounted for by excluding these children rather than by adjustment. Finally, sensitivity analyses were also conducted in the ESTELLE study by excluding the cases whose mother did not have a landline at home since they could not have been selected as controls (the information was not available in the ESCALE study).

Results

The pooled analysis included 357 cases (174 from ESCALE and 183 from ESTELLE) and 1,783 controls (949 from ESCALE and 834 from ESTELLE) younger than six years old (Table 1). Among cases, MYCN was amplified in 17.9% (11% of the cases under 18 months of age and in 25% of the older cases), non-amplified in 75.6%, and non-informative in 6.4%.

Table 1 Characteristics of the cases and controls of the ESCALE and ESTELLE studies
Full size table

Case–control comparability

In both studies the cases were more likely to be younger, first born, to live in an urban area, and to have younger mothers than the controls (Table 1).

Between-study heterogeneity

Control mothers in the ESTELLE study lived more often in less populated areas and were more highly educated than those in the ESCALE study (Table 1). They were also more likely to have used pesticides at home (p < 0.01) (results not tabulated).

Pesticides use

Overall, maternal use of any pesticide during pregnancy was reported for 43.7% of the cases and 35.7% of the controls. Insecticides were the most commonly used (40.6% of cases and 33.9% of the controls) and they were mostly used alone. Their use was mainly reported indoors (80.0% of insecticides use in cases and 81.0% in controls). Mothers rarely used herbicides or fungicides and they often also used insecticides.

The maternal use of any type of pesticide during pregnancy was associated with the risk of NB (OR 1.5 [95% CI 1.2–1.9]). There was a positive association with the use of insecticides alone (OR 1.4 [95% CI 1.1–1.9]) or insecticides with other pesticides (OR 2.0 [95% CI 1.1–3.4]). There was no between-study heterogeneity except for herbicide use (pooled OR 2.0 [95% CI 1.1–3.7]); ESCALE (OR 3.8 [95% CI 1.8–8.0]); ESTELLE (OR 1.1 [95% CI 0.4–3.2]); p value for interaction = 0.07).

Maternal occupational pesticide exposure during pregnancy was associated with the risk of NB (OR 2.0 [95% CI 1.0–4.0), although the frequency of exposure was low (3.6% of cases and 1.8% of controls) (data not tabulated).

The results were similar whether or not MYCN was amplified (Table 2), and when the analyses were stratified by age at diagnosis (Table 3). Among the controls, the prevalence of pesticide use varied by urban/rural status, but not by maternal level of education, and the results did not change when stratification for either of these factors was used instead of adjustment (Supplementary Table 1); neither did the exclusion of children with congenital malformations change the results.

Table 2 Maternal use of pesticides during pregnancy and risk of neuroblastoma
Full size table
Table 3 Maternal use of pesticides during pregnancy and risk of neuroblastoma by age at diagnosis
Full size table

Maternal use of pesticides before and after pregnancy and the use of professional pest control treatments at home were not collected in the ESCALE study. In the ESTELLE study, more than a quarter of the mothers reported pesticide use during all three time periods (29.0% of the cases and 27.6% of the controls) and very few mothers reported the use during pregnancy only (6.0% of cases and 3.5% of controls). This precluded specific analyses by time window.

No association was found with professional pest control treatments at home during pregnancy (OR 1.2 [95% CI 0.5–2.8]) (data not tabulated).

Finally, in the ESTELLE study, the results were unchanged in sensitivity analysis excluding the cases with no telephone landline.

Discussion

Our findings suggest that the maternal use of pesticides during pregnancy may be associated with an increased risk of NB.

In a large US case–control study, Daniels et al. [13] also reported modest associations (OR around 1.5) with the use of pesticides at home or in the garden during the preconception-pregnancy period or childhood. Consistent with our findings, this study showed similar estimates irrespective of the MYCN status, which does not support the potential for pesticides to act through different pathways in the two subtypes. In that study, the estimates were stronger among older children. Since the use of pesticides at home may be associated with life-patterns, the stronger associations observed in older children may reflect the effect of longer period of exposures to pesticides. Previous studies have also suggested that different etiologic factors may be specific to age at NB diagnosis [14, 20, 21]. Our study found no difference by age for the most prevalent exposures, pesticides and specifically insecticides. While there was a suggestion of differences by age with the use of herbicides and fungicides, this could be a chance finding since based on small numbers, the confidence intervals in the two groups overlapped substantially.

Another study conducted in Germany by Schüz et al. [14] found an association between NB and the use of household pesticides after the child’s birth, but did not investigate pesticide exposure during pregnancy.

Consistent with our findings, associations between NB and maternal occupational exposure to pesticides were reported in two case–control studies in the USA [15, 16]. In a cohort study in Norway, an increased risk of NB was observed in offspring of parents having worked in field vegetable [18]. However, these findings should be interpreted with caution since they are based on small numbers. In two of these studies [16, 18], like in ours, estimates were based on less than ten exposed cases and could represent a chance finding. Literature on paternal occupational exposure to pesticides is not supportive of an association with NB as shown in a meta-analysis conducted by Moore et al. [17]. A large Texan case–control study estimated residential exposures to pesticides due to neighboring agricultural activities and found no association [19]. However, estimates were inconsistent between low and high level of exposure and were based on less than 15 exposed cases.

It is biologically plausible that maternal pesticide exposure during pregnancy could be associated with the risk of NB. Pregnancy represents a critical window of exposure since some explanatory hypotheses suggest that NB is initiated in utero during the sympathoadrenal development from the neural crest [6]. It has been shown that maternal exposure during pregnancy can lead to fetal exposure since these compounds pass through the placenta and can be found in cord blood, infant hair, and meconium [11, 22]. The potential underlying mechanisms are still unknown. Many pesticides are suspected to have different mutagenic or immunotoxic properties and some individual pesticides have been classed as “probable or possible carcinogens” by the International Agency for Research on Cancer [9, 10]. In addition, because of the similarity of brain biochemistry, some insecticides that target the nervous system of insects may also be neurotoxic to humans [23]. However, in our study, because the use of pesticides was correlated between time periods (preconception, pregnancy, and childhood) we cannot identify pregnancy as a true critical period of exposure. It is possible that patterns of exposure are related to lifetime habits, which are normally consistent throughout the periconceptional period and later in childhood.

A limitation of our study is that the majority of mothers who reported any pesticide use (94%) actually used insecticides, either alone or combined with herbicides or fungicides, which limited our ability to investigate associations with pesticides other than insecticides. In addition, we could not identify the active ingredients in the products used as in our study we only asked women about the category of pesticide as we thought this would be recalled with greater accuracy than the actual product name. Commercial household pesticides often contain multiple active ingredients, all which may have different properties including potentially carcinogenic actions. In our study we mainly focused on maternal use of pesticides at home. However, the mothers may be exposed to other direct or indirect sources of pesticides which we did not account for, like the paternal use of household pesticides or other sources of environmental exposure.

As expected, the proportion of cases ineligible because they had died or were receiving palliative care was smaller under 18 months of age than for those aged 18 months or more. This may have introduced a selection bias if age was related to opportunity of pesticide exposure. However, the age-stratified analyses did not provide evidence of substantial differences in the associations between NB and maternal use of pesticides. As no information on exposure was available on cases and controls that refused to participate, we do not know if these children were comparable to the study sample. To overcome these limitations, we stratified our analysis for different factors that may have been related to control participation to limit selection bias.

Finally, the study of pesticides exposure relied on maternal self-report, which may involve both non-differential and differential bias. Because of the particular distribution of NB (52.7% of the cases younger than 18 months) the lapse of time between the exposure time and the interview was short, which may have limited non-differential measurement error, and which may not be the case for other childhood cancers. In our study, trained interviewers using highly structured questionnaires conducted interviews with the aim of reducing potential differential misclassifications. Despite this, it cannot be excluded because case mothers may have tended to think more deeply about their possible exposures than control mothers. Previous studies have found consistency between self-reported pesticide treatments and pesticides concentrations in dust [24] and that agreement about pesticides exposure between parents did not differ by case–control status [13], suggesting no differential recall based on motivation of case parents. Furthermore, our findings stratified by urban status were similar, despite differences in the reported prevalence of pesticide use across these strata.

Our studies also have several strengths. The ESCALE and ESTELLE studies were designed to be pooled with uniformly defined exposures, which facilitated the pooling process, making this one of the largest investigations of NB at present. In these population-based studies, the control participation rate was high and the cases were identified from a nationwide cancer registry, which has a high degree of completeness.

In conclusion, this pooled analysis adds to the evidence of an association between NB and maternal use of pesticides during pregnancy, in household or occupation. Because data were obtained retrospectively by questionnaire, recall bias is possible, particularly for domestic use. Replication by other large epidemiological studies with different designs is important. However, until a better study design can be found, our findings add yet another reason why to advise pregnant women to limit pesticide exposure during the periconceptional period.