The study mainly showed that both maternal and paternal household use of pesticides were significantly associated with childhood AL and NHL, but not with HL. Maternal household insecticide use during pregnancy was significantly associated with both the ALL and AML types of AL but only, for lymphomas, with Burkitt lymphoma and mixed-cell HL. The associations with herbicide and fungicide use during pregnancy were less marked.
The size of the study provided sufficient statistical power for most of the associations under study. For HL, the smallest case group, the statistical power of the study to show ORs of 1.5 and 2.0 was equal to 56% and 95%, respectively, for exposure prevalence of 40% in controls. The study may have suffered from a lack of power for that disease subgroup, but significant results were nonetheless obtained for some subtypes of the disease.
The cases were identified through the data collection system of the French National Registry of Childhood Blood Malignancies which has a high degree of exhaustiveness (> 99% of AL diagnosed in mainland France) (Clavel et al. 2004
), making case selection at the identification stage unlikely. The case mothers’ participation rate was very high, about 90%. Noninclusion was attributed mainly to the child’s poor condition. However, there is no obvious reason for household pesticide exposure being related to the severity of the disease or short-term survival, particularly because the associations were similar for rural, semiurban, and urban places of residence, where health care may differ.
We randomly selected the controls from the overall population. The national telephone directory was used as the basis for random selection. Unlisted numbers were randomly generated to prevent selection of controls on socioeconomic category or related factors that might influence inclusion in the telephone directory. The quota sampling process successfully ensured that responding controls had the same distribution as the whole case group with regard to sex and age, and the same distribution as the overall population with regard to birth order and region, as shown by comparison with the French national perinatal surveys (Blondel et al. 1997
). Degree of urbanization was not available in these surveys. The refusals to take part could have been related to parental socioeconomic status or educational level, which appeared higher among controls than among HL cases and, to a lesser extent, AL cases. Maternal educational level was very similar in the control group and in the French population, but paternal educational level was higher in the control group (Blondel et al. 1997
). However, parental socioeconomic status and educational level were not associated with household use of pesticide in the control group. Moreover, the results were unchanged after adjustment for those variables. Pesticide use was more frequent among controls who lived in a house, and controls lived slightly more often in a house than did cases; but the estimates were similar when stratified by housing category, and adjustment for the type of housing did not modify the results. There was therefore no indication that the controls had been selected particularly on factors related to household pesticide exposure. A previous French hospital-based case–control study found lower rates of parental pesticide use among controls, about 20%, but the study was conducted in more urban areas (Menegaux et al. 2006
The case and control interviews were conducted in the same manner, by the same interviewers, using closed questions. Misclassifications are likely to have occurred because the exposures to pesticides were described retrospectively and reported by maternal interviews. Difficulty in recalling the exposures that occurred during pregnancy probably increases with age. Recall should therefore be greater for mothers of children < 2 years of age. Interestingly, the associations with AL were slightly more pronounced for that age group, in which misclassifications might be weaker, than for the older age groups. In addition, the questionnaire did not include details on frequency of use, chemicals used, or conditions of use, which might have specified the exposure but might also have been more difficult to recall. Paternal exposure was particularly subject to imprecision and inaccuracies, because it was collected from the mother. Some sources of exposures during childhood such as maternal use of pesticide after birth or use of pediculosis lotions were not elicited. It is therefore possible that some indirect childhood exposures may not have been detected, inducing additional misclassifications. Differential recall of past pesticide exposure by cases and controls cannot be excluded, because the general public is becoming increasingly aware of the potential toxicity and carcinogenicity of pesticides. However, it is difficult to evaluate the direction of the resulting bias, if it exists, because cases could be expected to have over- or under-reported the exposure depending on whether recognition of past exposure or guilt about that exposure took precedence. Moreover, the possibility of case mothers reporting their pesticide exposure more accurately than control mothers cannot be excluded. Thus, we were unable to formulate the hypothetical sensitivity and specificity values for maternal reporting of pesticide use among cases and controls that should be required for further sensitivity analyses (Rothman and Greenland 1998
; Trivers et al. 2006
). However, in this study, positive associations were observed mainly for AL and NHL and mainly for maternal insecticide use during pregnancy. If systematic overreporting by case mothers explained the results, overreporting would not be expected to depend on the disease.
“Not known” answers were considered informative data, because three response options were consistently proposed: “ever used,” “never used,” and “not known.” Thus, mothers who were not sure about use of a pesticide were expected to have answered “not known,” which should have limited potential misclassifications. Moreover, the results were supported by the sensitivity analysis, which showed persistent associations even when missing data were allocated the “never used” response for cases and the “ever used” response for controls.
Only a few known factors may have confounded the results. Degree of urbanization and type of housing were independently associated with household use of pesticides in the study, with more frequent use by the controls living in a house or in a rural area; but the results were very similar with or without adjustment for those factors considered at the time of conception or interview. The results remained unchanged after adjustment for family history of cancer, birth order, early infections, and breast-feeding. Each household pesticide use might be a confounder for the others. The strong correlations made it difficult to disentangle the various pesticide exposures. The use of combined variables and multivariate analyses suggested that paternal use of pesticide was likely to be confounded by maternal use. Because the study focused on intrauterine exposure due to maternal pesticide use, the role of exposure during childhood could not be elucidated. However, the fact that the association was observed for children < 2 years of age and did not get stronger for older children points to a greater role of exposures during pregnancy. Finally, farming and occupational exposures to pesticides were very uncommon in the study population and were not related to hematopoietic cancers. Occupational exposures were therefore not confounders. Moreover, excluding children whose parents worked in agriculture or whose mothers were occupationally exposed to pesticide did not change the results.
Fewer studies have addressed household exposure. With regard to maternal household exposure to pesticides during pregnancy, five of the seven published studies on childhood AL (Buckley et al. 1989
; Infante-Rivard et al. 1999
; Leiss and Savitz 1995
; Lowengart et al. 1987
; Ma et al. 2002
; Meinert et al. 2000
; Menegaux et al. 2006
) reported significant and positive ORs. Lowengart et al. (1987)
reported significant ORs of 3.2 and 9.0, respectively, for indoor and outdoor maternal use of pesticide during pregnancy or nursing. Infante-Rivard et al. (1999)
observed significant associations between ALL and parental use of indoor insecticides during pregnancy, indoor plant insecticides (OR = 2.0; 95% CI, 1.3–2.9), outdoor herbicides (OR = 1.8; 95% CI, 1.3–2.6), and products for trees (OR = 1.7; 95% CI, 1.1–2.6). They reported dose–response relationships with some outdoor pesticide uses. Menegaux et al. (2006)
reported ORs of 1.8 (95% CI, 1.2–2.8) and 2.5 (95% CI, 0.8–7.2) for maternal use of home insecticides and garden pesticides during pregnancy. Leiss and Savitz (1995)
reported ORs of 3.0 (95% CI, 1.6–5.7) and 1.1 (95% CI, 0.6–1.9) for parental hanging of pest strips in the home and parental garden treatment. Ma et al. (2002)
found positive dose–response relationships and reported ORs of 2.1 (95% CI, 1.3–3.5), 0.8 (95% CI, 0.4–1.4), and 1.6 (95% CI, 0.9–3.0) for AL and insecticides use, flea control product use, and herbicide use, respectively. In a study of AML, Buckley et al. (1989)
did not distinguish between indoor and outdoor maternal exposure to pesticide during pregnancy, and they reported nonsignificant ORs of 1.4 for less than one use per week and 0.9 for between one and two uses per week. However, the authors reported that eight cases and no control mothers were exposed to pesticides on most days during pregnancy. Meinert et al. (2000)
reported a nonsignificant OR of 1.4 for parental indoor use of insecticides. In general, it would appear that indoor pesticide use during pregnancy may be more strongly associated with AL than the products used for gardening. In two studies, gestational single exposures to pesticides applied by pest exterminators were investigated. ORs of 2.2 (95% CI, 1.0–4.8) (Ma et al. 2002
) and 0.4 (95% CI, 0.1–1.2) (Leiss and Savitz 1995
) were reported.
Only three studies have investigated the relationship between household exposure to pesticides and childhood lymphoma (Buckley et al. 2000
; Leiss and Savitz 1995
; Meinert et al. 2000
). Two studies addressed only NHL (Buckley et al. 2000
; Meinert et al. 2000
). The third investigated both NHL and HL (Leiss and Savitz 1995
). In all three studies, at least one household pesticide exposure was significantly associated with childhood lymphoma. Leiss and Savitz (1995)
reported an increased risk of lymphoma with specialist pest extermination exposure during early childhood (OR = 1.8; 95% CI, 1.1–2.9]). However, for lymphoma, they did not report the strong association with pest-strip exposure that they observed with AL. Meinert et al. (2000)
observed positive and significant ORs for NHL and maternal exposure to indoor pesticide use during pregnancy (OR = 3.7; 95% CI, 1.8–7.6) and specialist pest extermination exposure during pregnancy or childhood (OR = 2.6; 95% CI, 1.2–5.7), whereas for AL the latter OR was 1.3 (95% CI, 0.8–2.3). Meinert et al. (2000)
also found a significant dose–response trend with parental indoor insecticide use during childhood. Buckley et al. (2000)
reported ORs consistently greater than unity and a nearly significant (p
= 0.05) trend for maternal exposure to insecticides during pregnancy, and significant ORs for childhood exposure to pesticides (OR = 2.4; 95% CI, 1.4–4.0) and professional pest extermination around the home during pregnancy (OR = 3.0; 95% CI, 1.4–6.2).
In the present study, the associations were stronger for common B-cell ALL and AML than for T-cell ALL or mature B-cell ALL. Burkitt lymphoma was also more strongly associated with maternal pesticide use during pregnancy than the other NHL. With regard to HL, only the mixed-cell subtype was associated with exposure. It is noteworthy that the two types of lymphoma associated with maternal pesticide use during pregnancy are both Epstein-Barr virus–related lymphomas, which may suggest that some kind of interaction between pesticide exposure and susceptibility to viral lymphomagenesis might exist. Few other studies have investigated the association between childhood leukemia or lymphoma subtypes and household pesticide use. Relatively similar associations with ALL and acute non-lymphoblastic leukemia were reported in each study (Lowengart et al. 1987
; Ma et al. 2002
; Menegaux et al. 2006
). Buckley et al. (2000)
found quite similar positive associations for all subtypes of childhood NHL.
Overall, increased risks of childhood AL or lymphoma have been reported for children exposed to household pesticides. Intrauterine pesticide exposure seems to have been a little more consistently associated with childhood leukemia and lymphoma than childhood exposure.
In this study, as in all the published studies, differential misclassification bias cannot be ruled out. Obtaining accurate prospective exposure assessments thus remains crucial.