Exposure to professional pest control treatments and the risk of childhood acute lymphoblastic leukemia

Authors

  • Helen D. Bailey,

    Corresponding author
    1. Telethon Institute for Child Health Research, Centre for Child Health Research, University of Western Australia, Perth, Western Australia, Australia
    • Telethon Institute for Child Health Research, PO BOX 855, WA 6872, Australia
    Search for more papers by this author
    • Tel.: +61 8 9489 7922, Fax: +61 8 9489 7700

  • Bruce K. Armstrong,

    1. Sydney School of Public Health, University of Sydney, Camperdown, New South Wales, Australia
    Search for more papers by this author
  • Nicholas H. de Klerk,

    1. Telethon Institute for Child Health Research, Centre for Child Health Research, University of Western Australia, Perth, Western Australia, Australia
    Search for more papers by this author
  • Lin Fritschi,

    1. Western Australian Institute for Medical Research, University of Western Australia, Perth, Western Australia, Australia
    Search for more papers by this author
  • John Attia,

    1. School of Medicine and Public Health, Centre for Clinical Epidemiology and Biostatistics, University of Newcastle, Newcastle, New South Wales, Australia
    2. Department of Medicine, John Hunter Hospital and Hunter Medical Research Institute, New Lambton, New South Wales, Australia
    Search for more papers by this author
  • Rodney J. Scott,

    1. Hunter Medical Research Institute, The School of Biomedical Sciences, Faculty of Health, University of Newcastle
    2. Hunter Area Pathology Service, HNEHealth, Newcastle, New South Wales, Australia
    Search for more papers by this author
  • Elizabeth Smibert,

    1. Children's Cancer Centre, Royal Children's Hospital, Melbourne, Victoria, Australia
    Search for more papers by this author
  • Elizabeth Milne,

    1. Telethon Institute for Child Health Research, Centre for Child Health Research, University of Western Australia, Perth, Western Australia, Australia
    Search for more papers by this author
  • for the Aus-ALL Consortium


  • The Aus-ALL consortium conducted the study and the Telethon Institute for Child Health Research (TICHR), University of Western Australia, was the coordinating centre. Bruce K. Armstrong (Sydney School of Public Health), Elizabeth Milne (TICHR), Frank M. van Bockxmeer (Royal Perth Hospital), Michelle Haber (Children's Cancer Institute Australia), Rodney J. Scott (University of Newcastle), John Attia (University of Newcastle), Murray D. Norris (Children's Cancer Institute Australia), Carol Bower (TICHR), Nicholas H. de Klerk (TICHR), Lin Fritschi (WA Institute for Medical Research), Ursula R. Kees (TICHR), Margaret Miller (Edith Cowan University), Judith R. Thompson (WA Cancer Registry) were the research investigators and Helen D. Bailey (TICHR) was the project coordinator. The clinical investigators were: Frank Alvaro (John Hunter Hospital, Newcastle); Catherine Cole (Princess Margaret Hospital for Children, Perth); Luciano Dalla Pozza (Children's Hospital at Westmead, Sydney); John Daubenton (Royal Hobart Hospital, Hobart); Peter Downie (Monash Medical Centre, Melbourne); Liane Lockwood, (Royal Children's Hospital, Brisbane); Maria Kirby (Women's and Children's Hospital, Adelaide); Glenn Marshall (Sydney Children's Hospital, Sydney); Elizabeth Smibert (Royal Children's Hospital, Melbourne); Ram Suppiah, (previously Mater Children's Hospital, Brisbane).

Abstract

Previous studies suggest that exposure to pesticides increases the risk of childhood acute lymphoblastic leukemia (ALL). The aim of this analysis was to investigate whether professional pest treatments in or around the home before birth or during childhood increased the risk of childhood ALL. Data from 388 cases and 870 frequency-matched controls were analyzed using unconditional logistic regression, adjusting for study matching variables and potential confounders, to calculate odds ratios (ORs). A meta-analysis of our findings with the published findings of previous studies was also conducted. The ORs for any professional pest control treatments were 1.19 (95% CI 0.83, 1.69) in the year before pregnancy, 1.30 (95% CI 0.86, 1.97) during pregnancy and 1.24 (95% CI 0.93, 1.65) for those done after the child's birth. The ORs for exposure after birth were highest when it occurred between the ages of two and three years. ORs were elevated for termite treatments before birth. ORs were higher for pre-B than T cell ALL and for t(12;21) (ETV6-Runx-1) than other cytogenetic sub-types. The pooled OR from a meta-analysis of our study with three previous studies of professional pest control treatments during pregnancy was 1.37 (95% CI 1.00, 1.88). Our results, and those of our meta-analysis, provide some evidence of a modestly increased risk of ALL for professional pest control treatments done during the index pregnancy and possibly in the child's early years. The analysis of pooled data from international collaborations may provide more certainty regarding these potentially important associations.

Acute lymphoblastic leukemia (ALL) is the most common type of childhood cancer in developed countries. In Western countries, the age standardized incidence rates are ∼30–40 per million.1 It is more common in boys and the majority of cases are diagnosed before 5 years of age.2 Little is known with any certainty about the causes of ALL, although it is likely that both genetic and environmental factors play a role.3 The early age at diagnosis suggests that parental (preconception) and fetal environmental exposures, as well as those in early childhood, may figure in disease development.

It is feasible that exposure of the parents or child to pesticides in key time periods could increase the risk of childhood ALL, as some pesticides have been identified as “possible” or “probable” carcinogens.4 Exposure of the mother or father before the pregnancy could result in damage to the reproductive cells. During pregnancy, exposure to pesticides has been demonstrated as their residual presence has been detected in umbilical cord blood,5, 6 neonatal hair and meconium.6 Children have higher respiratory and metabolic rates than adults and tend to play close to the ground with frequent mouthing of fingers, toys and other objects.7 For these and other reasons, they may be exposed to higher doses of environmental toxin, such as pesticides, than adults.

Previous studies have reported an increased risk of ALL with exposure to pesticides in the home or garden before pregnancy,8 during pregnancy8–15 or after the birth of the index child.8–10, 15–17 In addition, there is evidence linking occupational exposure to pesticides with childhood ALL.18 There are conflicting reports about whether professional pest control treatments in and around the home are associated with ALL. Three studies have suggested an association between using professional pest controllers around the time of the pregnancy,8, 9, 11 and three with their use after the birth of the child,8, 9, 19 others have not.16, 20 Several of these studies have lacked precision because of small numbers of exposed subjects.9, 19, 20

The Australian Study of Causes of Acute Lymphoblastic Leukaemia in Children (Aus-ALL) began in 2003 and aimed to investigate genetic, dietary and environmental causes of childhood ALL.21–24 The aim of this analysis was to investigate whether exposure to professional pest control treatments around the house and garden in specific time periods increased the risk of childhood ALL. We chose to investigate exposure to professional pest control treatments as our measure of home pesticide exposure because we thought it could be more accurately recalled by parents than any general pesticide use around the home or garden. In addition, greater applications of pesticide would be expected in professional treatments than in an average householder treatment. We also aimed to investigate whether the relationship varied by the location of the treatments, the type of pest being treated or by the frequency of treatments. Throughout this article, we use the term pest control treatments to describe treatments for pests such as insects, spiders, rodents and birds; treatments with herbicides are not included.

Abbreviations

ACT: Australian Capital Territory; ALL: Acute lymphoblastic leukemia; Aus-ALL: Australian Study of Causes of Acute Lymphoblastic Leukaemia in Children; CATI: computer assisted telephone interview; CI: confidence interval; NSW: New South Wales; NT: Northern Territory; OR: odds ratio; QLD: Queensland; RDD: random digit dialing; SA: South Australia; TAS: Tasmania; VIC: Victoria; WA: Western Australia

Material and Methods

Study population

Aus-ALL was a national, population-based, case-control study of childhood ALL conducted in Australia between 2003 and 2007, which recruited 416 cases and 1361 controls aged younger than 15 years. For a child to be eligible, the biological mother needed to be available and have sufficient English skills to complete the questionnaires. Case families were identified and recruited through all ten pediatric oncology centers in Australia. Cases were eligible if they had been diagnosed between 1st July 2003 and 31st December 2006 and had reached remission. The study had Human Research Ethics Committee approval from all participating hospitals. Controls were recruited by random digit dialing (RDD) in seven recruitment waves between 2003 and 2006 and were frequency matched to cases by age, sex and State of residence in a ratio of ∼3:1. Full details of the study population, including a full description of the control recruitment methods, have been published.21–24

Data collection

Both parents were mailed questionnaires that included questions about a range of potentially leukemogenic exposures. The mothers' questionnaire contained questions about professional pest control treatments in or around the home or garden. Mothers were asked if the pest control had been done in three time periods: the year before the index pregnancy, during the pregnancy, and since birth of the index child.

Once the mother's questionnaire was returned, she completed a computer assisted telephone interview (CATI) to collect further data on exposures listed in the questionnaire. In the CATI, data were collected about the frequency of professional pest control treatments and, for each individual episode, the pests they were trying to eliminate and whether it was done inside or outside the home. If the treatment was done during the pregnancy, information was sought about the trimester. We asked whether either parent was home when the treatment was done before the pregnancy, the mother was home when treatments were done during the pregnancy, and whether the child was home when treatments were done since birth. We also asked whether the bedroom of the person(s) of interest was treated and whether the person slept in it within 3 days of the treatment being done.

The fathers also completed a CATI. If the father said he had lived at a separate address from the mother in the year before pregnancy, he was asked to provide information about any professional pest control treatments.

Exposure metrics

We assessed whether the association between ALL and professional pest control treatments varied by the timing of the exposure, the location or frequency of the treatment, or the type of pest being treated.

Separate analyses were done for the year before the index pregnancy, during the pregnancy, and from birth until the censoring date. For cases, the censoring date for exposure was the date of diagnosis. For controls, it was the date when the self-administered questionnaire was returned; because controls were frequency matched, there was no directly corresponding case whose date of diagnosis could have been used. We analyzed data for the period from the year before pregnancy until the censoring date so that we could compare our findings with those of other studies. In addition, we analyzed whether the association between ALL and pest control treatments after the child's birth differed by the child's age at the time of the treatment. The age groups used for these analyses were based on those recommended by the US Environmental Protection Agency.25

The options for location of the pest control treatments were “inside the house,” “outside the house” or “both.” The final categories used were “any pest control inside” and “any pest control outside,” as few people had the treatments done only inside or only outside.

To estimate the “dose” of pest control treatments during childhood, we categorized the total number of treatments into three groups (1, 2, > 2).

The pests treated were grouped into three categories: termites; general insects and spiders; and rodents or birds.

Immunophenotype and cytogenetic classification

We obtained information about immunophenotype and cytogenetic sub-types of participating cases from the medical record details provided by clinicians. The latter were determined using fluorescence in situ hybridization screening.

Statistical analysis

Odds ratios (OR) and 95 percent confidence intervals (95% CIs) were estimated using unconditional logistic regression in SPSS for Windows version 15, released Sept 2006 (SPSS, Chicago IL.). All models were adjusted for the study matching variables—age at diagnosis, sex and State of residence. Variables considered a priori to be potential confounders of the association between the use of professional pest controllers and risk of ALL (ethnicity, maternal age at the child's birth, birth order, family income, highest parental education and the child having a birth defect) were assessed for inclusion in the models. Variables that met empirical criteria for potential confounding (that is, were independently associated with both pest control treatments and risk of ALL) were included in the final models. The final models contained the matching variables and maternal age at the child's birth, birth order, parental education and income.

The reference category for the analyses was no professional pest control treatment from the year before pregnancy until the censoring date. For the analyses of exposure by the child's age when the treatment was done, only unexposed children who had reached the mid-point of the age group were included in the analyses. For example, only unexposed children who were 6 months or older were included in the analysis of treatments done before the age of 1 year.

We also analyzed the data by immunophenotype and cytogenetic subtypes for types with more than 50 cases.

Meta-analysis

We conducted a meta-analysis of our findings with the published findings of previous studies. To be included in the meta-analysis, each study was required to: (i) be a cohort or case-control study which presented ORs and corresponding 95% CIs for the association between professional pest control treatments before or during pregnancy or after the birth of the child and risk of childhood ALL (or provided data that allowed them to be calculated); and (ii) be population-based and not restricted to any specific population subgroup (e.g., children with Down Syndrome). We searched PubMed for original studies of pesticide use and risk of childhood ALL, published from 1966 to 2009 using leukemia, pesticide and child as search terms. All journal articles about domestic pesticide use were read to identify studies that specifically investigated the use of professional pest controllers.

Six case-control studies were identified;8, 9, 11, 16, 19, 20 one was excluded from all meta-analyses because it was restricted to children with Down Syndrome.11 Some studies were included in the meta-analyses although they investigated slightly different time periods of exposure than our study did (Supporting Information Figure S1).

One study was included in the meta-analysis for pest control treatments before pregnancy, despite including only exposures in the three months before pregnancy.8 Three studies were eligible for inclusion in the meta-analysis for pest control treatments during pregnancy,8, 9, 16 including one that only included exposures in the third trimester16; one for any pest control treatments from birth until diagnosis19; and one for pest control treatments in the first few years of life.8 Two studies were included in the meta-analysis of treatments at any time from before pregnancy until early childhood, despite slightly different time periods before pregnancy or after birth.8, 20 The two separate studies by Meinert et al.19, 20 were included despite using case definitions of “acute childhood leukemia” and “childhood leukemia,” respectively, as ALL accounts for 75–80% of these groups.2 One study used similar categories of pests to us9 and was included in a meta-analysis of subgroups of types of pests. We extracted the most appropriate OR from each study, and used fixed effects, precision-based weighting26 to calculate summary ORs with the results from Aus-ALL. Statistical heterogeneity among studies was assessed using the Cochrane Q test. Forest plots were produced using STATA version 10, released June 2007 (StataCorp, College Station, TX).

Results

We were notified of 568 incident cases of ALL, of whom 49 were ineligible to participate: 30 from non-English speaking backgrounds, 12 overseas visitors, three whose biological mother was unavailable, and four who did not reach remission, leaving 519 eligible cases. Parents of 416 (80.2%) cases consented to participate in the study, and 388 mothers (74.8% of eligible) returned the questionnaire and of these 384 (99.0%) also did the telephone interview (Table 1). The median time between diagnosis and remission was 35 days and between diagnosis and return of the questionnaire, 7 months.

Table 1. Demographic characteristics of cases and controls
inline image

Of the 2,947 known eligible control families identified through RDD, 2,071 (70.3%) agreed to take part (participation was 54.7% when families who's eligibility could not be determined were treated as eligible21). Because of age and sex quotas for frequency-matching controls to cases, only 1,361 of these families were actually recruited into the study and sent questionnaires. Of the recruited mothers, 870 (63.9%) returned the questionnaire and of these 852 (97.9%) also did the telephone interview (Table 1).

The demographic characteristics of cases and controls who returned the written questionnaire were generally similar (Table 1). Case children, however, were more likely to be a first born child (47.7%) than were control children (41.6%) or to have a birth defect (5.7% versus 3.1% respectively). Control parents were more likely than case parents to be tertiary educated, have a higher income and to have been aged 35 years or older when the child was born.

Exposure to pest control treatment

The frequencies of the exposure metrics among cases and controls are shown in Table 2. There was weak evidence of an association with childhood ALL of professional pest control treatments in the year before the pregnancy (OR 1.19, 95% CI 0.83, 1.69), during the pregnancy (OR 1.30 95% CI 0.86, 1.97) or after birth (OR 1.24, 95% CI 0.93, 1.65) (Table 3). The OR for having pest treatments done in the first trimester of pregnancy was 1.25 (95% CI 0.54, 2.88), in the second trimester, it was 0.95 (95% CI 0.47, 1.93) and in the third, it was 1.98 (95% CI 1.00, 3.92) (results not tabulated). As most people who had pest treatments in the year before or during the pregnancy had only one treatment in each period, it was not possible to investigate whether there was a dose response. There was little evidence of increasing risk with increasing frequency of pest treatments after birth (data not shown). The ORs for having pest treatments done were highest if the treatments were done when the child was between two and three years of age 1.70 (95% CI 1.15, 2.52) (Table 4). The ORs were not materially different when the analyses were based on the age at first treatment, except that the OR for 6–<11 years fell substantially but was imprecise because of small numbers of children first exposed in this age group (OR 0.53 95% CI 0.16, 1.84; other results not shown).

Table 2. Frequencies of exposure to pesticides among cases and controls
inline image
Table 3. Odds ratios and 95% confidence intervals for risk of ALL with exposure to professional pest control treatments around the home and garden in relevant time periods
inline image
Table 4. Odds ratios for exposure to professional pest control treatments around the home and garden after the child was born: by the child's age when the treatment was done. Children with multiple exposures can appear in multiple age at exposure categories
inline image

The ORs for any treatments inside the house were similar to those for any treatments outside for each time period (data not shown), but most treated homes were treated both inside and outside.

The majority of pest control treatments in all time periods were for general insects and spiders and there was little evidence of any increased risk (Table 3). The ORs for termite treatments before and during pregnancy were elevated, although the numbers were small and the CIs wide (Table 3). Very few people had treatments for rodents or birds so no ORs were estimated.

For treatments in the year before pregnancy and during the pregnancy, there was no evidence of any greater increase in risk if the mother was home when the treatment was done or if her bedroom was treated (data not shown). Similarly, there was no evidence of an increased risk if the father was home or if his bedroom was treated before the pregnancy, or if the child was home or his/her bedroom was treated (data not shown).

When the analyses were done by immunophenotypes, there was evidence of increased risk of Pre-B cell ALL for pest control treatments in any time period. This was not the case for T cell ALL, where the OR estimates were all below unity but imprecise as there were few cases (Table 5). There was also evidence of an increased risk among cases with t(12;21) translocations (ETV6-Runx-1) with any exposure (Table 5). Similarly, the OR for any termite treatments in the year before or during the pregnancy was greater than two among cases with Pre-B cell ALL, t(12;21) translocations or with a normal karyotype (Results not shown). These estimates were imprecise because of small numbers of exposed cases.

Table 5. Odds ratios for exposure to professional pest control treatments around the home and garden by immunophenotype and cytogenetic subtype1
inline image

When all the analyses were repeated excluding all those with a birth defect, rather than adjusting for a birth defect, there was little change in the results (data not shown).

Meta-analyses

Details of previous studies included in the meta-analysis are summarized in Supporting Information Table S1 and the pooled estimates are shown in Figure 1. The ORs for the two studies of pest treatments before pregnancy were both above unity and the pooled OR was 1.26 (95% CI 0.90, 1.76). The ORs from three of the four studies of pest treatments during pregnancy were above unity, and the pooled OR was 1.37 (95% CI 1.00, 1.88). The results for the two studies of treatments done between birth and diagnosis were also similar, and the pooled OR was 1.25 (95% CI 0.97, 1.61). The pooled ORs for treatments done in the first, second and third years of the child's life were 1.39 (95% CI 0.94, 2.04), 1.41 (95% CI 0.98, 2.02) and 1.78 (95% CI 1.25, 2.52), respectively. The pooled OR for any treatment from a year before pregnancy through early childhood was 1.37 (95% CI 1.08, 1.75). The pooled OR for any termite treatments from the year before the pregnancy until diagnosis was 1.17 (95% CI 0.71, 1.93) and for any treatments for general insects in the same time period was 1.34 (95% CI 1.06, 1.70). The measure of heterogeneity (I2) for all analyses were low to moderate, except for pest control treatments done in the child's second year of life (I2 = 78.6%, p = 0.031). This result should be viewed with caution and it is only included in the table for completeness.

Figure 1.

Forest plot showing study-specific and summary OR for exposure to pest control treatments in different time periods. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Discussion

Our study showed a weak relationship between having the home or garden professionally treated for pests in the year before pregnancy, during pregnancy and after the child's birth and the risk of childhood ALL. There were suggestions that the risk was higher for treatments done when the child was between 2 and 3 years of age, or for termite treatments done in the year before or during the pregnancy. These results are consistent with the results of the meta-analyses, based on up to six studies including Aus-ALL.

As can be seen in Figure 1, our results are consistent with those of three8, 9, 19 of the five previous studies that have investigated the association between professional pest treatments and leukemia that were included in the meta-analysis.8, 9, 16, 19, 20 In addition, our results are also consistent with the only study about professional pest control treatments that was not included in the meta-analysis.11 This US study was restricted to children with Down Syndrome11 and reported ORs of 2.25 (95% CI 1.13, 4.49) for exposure in the time between one month before pregnancy and the end of pregnancy.

Only two studies16, 20 have not found an association with pest control treatments and leukemia. The first of these20 may not have observed any association with pest control treatments because the prevalence of pest control treatments was less than 2% in their German control population. This is a much lower prevalence than in our control population (>30% of controls) or that reported in several US studies.8, 11 The remaining study that did not find any association16 included only 73 cases of “leukemia,” 59 of which were ALL.27 There could have been differential misclassification of exposure as the cases were recruited up to 9 years after diagnosis and controls were restricted to children who had lived in their current residence at the time their matched case was diagnosed.

There is also evidence that the use of pesticides other than by professional pest controllers is associated with childhood ALL. A recent meta-analysis concluded that risk of childhood ALL was associated with unspecified residential pesticide exposure during pregnancy (summary OR 1.54, 95% CI 1.13, 2.11) and during childhood (summary OR 1.61, 95% CI 1.33, 1.95).28 A review of parental occupational exposure concluded that, while there is evidence of an increased risk with maternal prenatal occupational exposure to pesticides (OR 2.08, 95% CI 1.51, 2.88), there is less consistent evidence for paternal occupational exposure.18

Regarding the higher risk for toddlers, it is plausible that young children receive relatively high doses of pesticides because of hand to mouth behavior in the toddler years, which decreases after the age of six.29 After handling contaminated surfaces and objects, they may ingest or absorb any pesticide residues through the skin. Pesticide residues have been detected on household surfaces and toys for at least 2 weeks post application.30

We found a higher risk associated with pest control treatments for termites than with those for general insects and spiders. The only other study that investigated the types of pests being treated did not.9 However, because of the low prevalence of termite treatments in Quebec (∼1% of controls), where the study was done, results were reported only from the year before pregnancy until diagnosis, for which there was no evidence of an association. In our study, the prevalence of termite treatments in controls was higher (4.0% in the year before pregnancy, 2.9% during pregnancy and 9.0% after birth); thus, we were able to investigate the relationship by time period.

In Australia, chemical treatments for termites can only be done by licensed pest controllers.31 There are a limited number of chemicals approved for termite treatment and these include deltamethrin, bifenthrin and permethrin, which are all pyrethroids; chlorpyrifos, which is an organophosphate; fipronil which is a phenylpyrazole; and imidacloprid, which is a chloronicotinyl.32 At least three of these chemicals (bifenthrin, permethrin and fibronil) are classed as “Group C—Possible human carcinogens” by the US Environmental Protection Agency.33 There is a much wider range of chemicals available to treat other household pests such as spiders and ants. Pyrethroids are popular because they are effective against a range of pests, including termites in some cases,34 although a range of other pesticides can also be used. The toxicity of individual pesticides varies. As information about the actual pesticides used was not available, we were unable to determine which were associated with the increased risk.

Ours is the first study that has analyzed the risk of professional pest control treatments by immunophenotype and cytogenetic subtypes. We found that professional pest control treatments in the year before pregnancy, during pregnancy or after birth appeared to increase the risk of the t(12;21) (ETV6-Runx-1) subtype of ALL. About 25% of children with Pre-B cell ALL have t(12;21) translocations.35 There is very limited epidemiologic evidence of exposures associated with the presence of t(12;21) translocations. Scelo et al.36 reported that both prenatal and postnatal exposure to paint, but not solvents, was associated with an increased risk of the t(12;21) (ETV6-Runx-1) ALL subtype. In the analysis of house painting exposure data in Aus-ALL, we also found some evidence of house painting at any time from the year before pregnancy until diagnosis was associated with an increased risk of the t(12;21) (ETV6-Runx-1), but this finding was based on small numbers.21 It is possible that exposure to a variety of chemicals in the environment, including those in paints and pesticides is one of several causes of t(12;21) translocations. Both our study and that of Scelo et al.36 found the increased risk was not restricted to those exposed in the prenatal period.

Aus-ALL has strengths and limitations. Cases were ascertained from oncology centers that treated virtually all children with ALL in Australia and 75% of parents of eligible children agreed to participate; 70% of eligible controls identified by RDD also agreed to participate. However, 64% of participating control parents returned the questionnaires compared to 93% of participating case parents, raising the possibility of selection bias. We had information on parental education for all participants and on income for all but three cases and five controls. Control parents were more likely than case parents to have tertiary education and, using area-based measures, we have shown that Aus-ALL parents were of higher SES than the Australian general population.22 Among the controls in our study, the proportion of families who had pest control treatments was higher among families with higher income and those with at least one tertiary educated parent. Therefore, we may have overestimated the prevalence of pest control treatments in the Australian general population. SES was considered a priori to be a potential confounder of the association under study and both parental education and income in our models, but despite this, our ORs may underestimate the true magnitude of the association between pest control treatments and ALL.

For controls, the censoring date was the date the questionnaire was returned, while for cases it was the date of diagnosis. Controls were frequency matched by age to cases, but they had the opportunity to accrue exposures for the time period between study recruitment and return of the questionnaire as the questionnaire asked about exposures ‘since your child was born’ while cases were asked about exposures until diagnosis. The median time between control recruitment and return of the questionnaire was 44 days (inter-quartile range 42.5). It is unlikely that this short time would have had a significant impact on whether a child was exposed.

As with most case-control studies, recall bias is a possibility. We attempted to minimize this by using standardized written questionnaires and structured telephone interviews, with the interviewers blinded to case/control status. These methods would not prevent case parents potentially thinking more deeply about past exposures, and therefore reporting them more frequently. As reports about a possible association between the use of pesticides and ALL were published as far back as 1978,37 case parents in our study could have been aware of the possible association. A basic search of the internet yields thousands of references to this association. However, if rumination bias was an issue, similar associations would be expected across all time periods. This was not the case for termite treatments, where an association was observed with treatments done before and during pregnancy but not after birth. However, for any general insect and spider treatments, the ORs were lower and similar for all time periods.

Unlike some previous studies, we asked specifically about professional pest control treatments, rather than about the use of pesticides in general, in an effort to reduce measurement error, as it is reasonable to assume that mothers would be able to recall these events with greater accuracy. A limitation of this and previous studies is that we did not ask for the name of the actual chemical used in the treatments; it was considered unlikely the respondents would be able to provide this level of detail. Unfortunately there is no central database of chemicals used by professional pest controllers in Australian homes so that we could not do more specific analyses to identify both potentially harmful chemicals and those which are safe. Further research to develop biomarkers of past exposures to pesticides could also remove the need to rely solely on parent recall.

In conclusion, Aus-ALL provides some evidence of a modest increased risk of ALL associated with professional pest control treatments, particularly when the child was aged between two and three years, or if termite treatment was done before or during the index pregnancy. Our meta-analyses also provide some evidence of an increased risk of ALL associated with pest control treatments during the index pregnancy and in the early years of the child's life. Our findings add to the evidence that residential pesticide use during susceptible times is associated with the risk of childhood ALL. Further analyses using pooled data may provide more certainty to these potentially important findings. Such analyses are planned in the Childhood Leukemia International Consortium (see http://clic.berkeley.edu).

Acknowledgements

Aus-ALL was funded by The National Health and Medical Research Council (NHMRC) Grant 254539. Helen Bailey was supported by NHMRC Post Graduate Scholarship 513934, Elizabeth Milne by NHMRC Career Development Award 513910, and Lin Fritschi by NHMRC Fellowship 513706. Rodney Scott was supported in part by the Nine Broadcasting Network Children's Cancer Research Fund.

Ancillary