Main

Brain tumours are the second most frequent type of childhood cancer with a high mortality rate and a high frequency of long-term morbidity and psychosocial sequlae (Reimers et al, 2003). The annual incidence rates range from 20 to 40 cases per million children, with the highest rates reported in the Nordic countries (Peris-Bonet et al, 2006; Lannering et al, 2009; Schmidt et al, 2011). Better knowledge of the aetiology of childhood and adolescent brain tumour (CABT) for primary prevention strategies is therefore important.

A review of epidemiologic studies by Schüz and Kaatsch, 2002 suggests that infections could initiate or modify the risk of CABT similar to the infection hypothesis that has been proposed for childhood acute lymphoblastic leukaemia (Schüz et al, 1999; Roman et al, 2007; Kamper-Jorgensen et al, 2008). Exposure to infectious agents early in childhood could cause an abnormal immune response when exposed to infectious diseases later in childhood (Kinlen, 1995; Greaves, 2006).

Some studies have not found an association between infectious agents and CABT (Birch et al, 1990; Little, 1999; McKinney et al, 1999). Most of these studies, however, had small sample sizes and therefore low statistical power to detect an association and other studies have found an association between risk of CABT and various measures of infectious diseases (Linet et al, 1996; Dickinson et al, 2002; Shaw et al, 2006) or indications of an association with exposure to polyoma viruses (SV40, JCV and BKV) (Bunin, 2000; Vilchez and Butel, 2003; Khalili et al, 2003).

Children’s exposure to infectious diseases in early childhood is most commonly through contact with other children (Ma et al, 2009). Studies have shown that attending daycare increases a child’s risk of getting infectious diseases (Thrane et al, 2001; Kamper-Jorgensen et al, 2006; von Linstow et al, 2008), indicating that social contacts could reflect the general exposure to infections. Studies investigating the relationship between social contacts and CABT have, however, only found little or no evidence of an association (Shaw et al, 2006; Harding et al, 2009; Schmidt et al, 2010).

To evaluate the infection hypothesis for CABT further, we investigated the patterns of infectious diseases and social contacts in early life in relation to risk of developing a brain tumour during late childhood and adolescence in the CEFALO study.

Materials and methods

The CEFALO study is a multinational case–control study conducted in Denmark, Norway, Sweden and Switzerland, with the primary aim to investigate the possible association between CABT risk and mobile phone use and other relevant exposures (Aydin et al, 2011).

Case and control ascertainment

Eligible cases were children diagnosed with a primary intracranial brain tumour (Aydin et al, 2011) aged 7–19 years at the date of diagnosis. The study period was from 1 January 2004 until 1 August 2008, but varied slightly across the four countries (Denmark: January 2004–April 2008; Sweden: April 2004–August 2008; Switzerland: May 2004–2008; Norway: September 2004–August 2008).

To attain complete case ascertainment, cases were identified both at hospitals and in country-specific disease registries (the Danish National Cancer Registry (Tulinius et al, 1992; Storm et al, 1997); Danish Childhood Cancer Registry, Danish Pathology Registry, Danish National Patient Registry, the Swedish Regional Cancer Registries (Barlow et al, 2009), the Cancer Registry of Norway (Larsen et al, 2009) and Swiss Childhood Cancer Registry). Date of diagnosis was defined as date of first diagnostic imaging confirming a CABT.

Two controls per case were randomly selected from nationwide population registries in Scandinavia and in Switzerland from communal registries, and were individually matched by age (Denmark, Switzerland and Sweden: month of birth; Norway: year of birth), gender and region. Exposure of controls was censored at the date of diagnosis of the matched case.

Subjects diagnosed with Recklinghausen neurofibromatosis (12 cases) and tuberous sclerosis (one case) were excluded as their aetiology is genetic. Cases and controls with severe autism, severe mental retardation or complete deafness before the date of diagnosis were also excluded (two cases and two controls). In addition, families with insufficient language skills to complete an interview were excluded (15 cases and 36 controls). The study was approved by the National Data Protection Boards and ethical committees in all participating countries.

Data collection

Cases diagnosed before June 2006 (Norway: December 2007) and corresponding controls were included retrospectively. All other cases were included prospectively. A structured personal interview was performed with the child and one or preferably both parents by a trained interviewer during the time period 2006–2009. For ethical reasons, the interview was conducted at least 6 months after the date of diagnosis. Questionnaire translations were validated and pilot tested in all participating countries.

Variables and analysis strategy

We used the following variables as measures of social contacts and of exposure to infectious disease: living with other children before 6 years of age of index child (yes/no); birth order (1st born/>1st born); attending daycare in first year of life (yes/no) and before 6 years of age(yes/no); attending baby groups (at least once a week) within the first year of life (yes/no); mean number of days where child’s general condition was appreciably affected by infectious disease (once or less per month, 2–3 days per month, 4 days or more per month) at different ages (<1, 1–<3, 3–<6 years), calculated as a weighted average of answers provided per semester for the first 3 years of life and for the next 3 years of life combined (see Supplementary Appendix 1 online). The list of infectious diseases included: cold, fever without known cause, middle ear infection, tonsillitis, bronchitis, pneumonia, skin infection, urinary tract infection, stomach flu and others (e.g. whooping cough, 3-day fever, scarlet fever). For each child we also assessed cumulative daycare exposure by calculating total daycare childhours (categorised into quintiles) during the first 6 years of life as used by Ma et al (2002): Number of months attending a specific daycare × 4.35 (average number of weeks per month) × hours per week at daycare × number of children at daycare. The childhours were cumulated over all daycare facilities attended during the time period of interest. As early brain tumour-related symptoms might have influenced daycare attendance, all daycare information for the past 2 years before diagnosis or end of exposure date were disregarded (only relevant for the 7 and 8 year olds in our study). In addition, information on allergic conditions (yes/no) was available for all children, defined as having at any time before diagnosis of the matched case: doctors diagnosis of asthma, hay fever or eczema; or wheezing or whistling in the chest; itchy/watery eyes and sneezing/runny/blocked nose not in connection with a cold or the flu; or rash of folds of elbows, knees, front of ankles, under buttock or around neck, eyes or ears.

For statistical analysis, tumours were classified as: ‘Glioma’ (astrocytomas, other gliomas), ‘intracranial embryonal CNS tumours’ (including primitive neuroectodermal tumours (PNET)) and ‘Others’ (ependymoma, other specified intracranial neoplasms, unspecified intracranial neoplasms).

Data were analysed by conditional logistic regression in SAS (version 9.2; SAS Institute Inc., Cary, NC, USA). Relative risk estimates were expressed as odds ratios (OR) with associated 95% confidence intervals (CI). Highest attained education of parents as a proxy for socioeconomic status was included in the models as potential confounder, but did not alter the risk estimates and was therefore omitted from the final model (data not shown).

Results

In total, we invited 423 brain tumour cases and 909 controls and their families to participate in the CEFALO study. We interviewed 352 cases and 646 controls with participation rates of 83 and 71%, respectively. The most common reasons for non-participation are shown in Table 1.

Table 1 Characteristics of cases and controls
Full size table

The largest group of CABT in the study was glioma (55%) followed by ‘other tumours’ (28%) and intracranial embryonal tumours/PNET (18%) (Table 1). This corresponds to the overall distribution of tumour types in the Nordic countries (Schmidt et al, 2011).

Overall there were slightly more boys than girls in the study. Two-thirds of the participating children and adolescents were between age 7 and 14 years, while only one third was between 15 and 19 years (Table 1). A large proportion of children attended daycare before 6 years of age (87%). The majority of children attended baby groups (72%) and most (94%) of the index children were living together with siblings or other children. Regarding monthly number of days with infectious disease during the first 6 years of life, the majority of children were reported to have infectious disease once or less per month (67%), 20% 2–3 days per month and 9% 4 days or more per month.

Overall, no OR of any measure of social contacts was statistically significantly increased or decreased (Table 2); this did not change when stratifying by gender (data not shown). Furthermore, no changes in ORs were observed when controlling for allergy (never/ever allergy) (data not shown). However, a tendency of slightly increased risks were found for glioma and total childhours (only significant for fourth quintile; OR: 3.39; 95% confidence interval (CI): 1.54–7.46), living with other children before 6 years of age (OR: 1.64; 95% CI: 0.73–3.67) and attending baby groups during the first year of life (OR: 1.47; 95% CI: 0.95–2.28). Such a pattern was not observed for other CABT.

Table 2 Associations of attending daycare within 1st and 6 years of life, total daycare childhours within 6 years of life, living with other children, birth order and attending baby groups within 1st year and CABT
Full size table

With regard to infectious disease (Table 3), the OR for all CABT types combined was 1.27 (95% CI: 0.89–1.80) for children having 2–3 sick days per month in the 0–6 years of age and 1.87 (95% CI: 1.19–2.97) for children with 4 or more sick days per month (0–6 years) compared to children with one sick day or less per month (Table 3). Girls tended to have higher risk estimates for CABT than boys (data not shown). Controlling for diagnosed allergy did not alter the risk estimates. When stratifying by tumour type, increases were restricted to glioma and embryonal CNS tumours: an average of 4 or more monthly sick days at ages 0, 0–2, 3–5 or 0–5 years was associated with significantly increased risk of glioma. The observed risk was of similar magnitude, regardless of period of exposure, with the highest risk seen when looking at sick days over the first 6 years of life (OR: 2.93; 95% CI: 1.57–5.50). There was a tendency for similar increases for embryonal CNS tumours, the risk estimates tended to be slightly lower than for glioma, except when looking at the first 6 years of life combined where the OR was 4.21 (95% CI: 1.24–14.30), numbers were, however, small and the CI very wide.

Table 3 Self reported sick days from infectious disease and risk of CABT
Full size table

For other CABTs, no indication of a risk increase with self-reported number of sick days was observed (Table 3).

We tested for interaction between total daycare childhours and number of days of infections (0–<6 years), as well as for birth order and number of days of infections (0–<6 years), but no interactions were found.

Discussion

Overall, we found little evidence for an association between early-life social contacts and CABT risk. There was no evidence of an exposure–response effect of higher CABT risk for cumulative childhours of attending daycare facilities. In all, confidence intervals for all social contact measures were relatively wide and no systematic patterns were observed. However, we found some evidence for an association between number of sick days due to infectious disease and CABT risk. When subdividing by tumour type, the effect was restricted to glioma and embryonal CNS tumours, with the results showing statistically significant positive associations for sick days at all ages.

Our findings of no increased CABT risk in relation to social contacts are broadly consistent with earlier studies, which showed only small or no associations (Shaw et al, 2006; Harding et al, 2009; Schmidt et al, 2010). The overall association of CABT with sick days because of infections found in this study is in agreement with some other studies; however, the majority of studies did not find an association (Little, 1999) and, in addition, no specific infectious agent had been identified (Schüz and Kaatsch, 2002).

The observed association between CABT and infections could result from a common underlying cause, for example, an immunological factor that increases both risk of tumour development and susceptibility to infectious disease. Also, we cannot rule out that inverse causality could be an explanation, as having a yet undiagnosed brain tumour could make the case children more susceptible to infections. Both of these explanations would fit with the fact that we observed no clear association with particular age periods when the infections occurred.

We estimated infections in the first 6 years of life from self-report (infectious calendar developed for the purpose of this study, see Supplementary Appendix 1 online). Accurate recall of infectious episodes is, however, demanding (Schmidt et al, 2010) and laypersons may have difficulty in separating sick days due to infection from sick days for other reasons. Furthermore, case parents may have over-reported exposure, in an attempt to provide an explanation for the disease of their child; if so, one might expect them to do this for all age periods in the questionnaire, but inspection of the data for the most sick children showed no clear evidence of such an effect (data not shown). Alternatively, one could speculate that case parents might under-report sick days, as infections may be trivial compared with the later burden of having a child with a cancer diagnosis. In addition, common early symptoms of CABT include headache, nausea and vomiting (Wilne et al, 2007), and there may be a diagnostic delay of several years (ng-Tan and Franco, 2007, 2008; Wilne et al, 2007; Raab and Gartner, 2009) before the child is diagnosed with CABT; therefore, it is possible some reported infections of cases could reflect prodromal symptoms, even though we excluded any exposure occurring within 2 years of diagnosis. Altogether, the above factors may have introduced differential information bias. In this case we would, however, expect to find the same effect for all subtypes of CABT, whereas we only saw elevated risk estimates for glioma and embryonal CNS tumours, suggesting that biased reporting of infectious diseases is not a major concern. On the other hand, the group of other CABTs is small and the difference may be due to random variation.

Another concern is if first born children may have had more sick days due to more concerned first time parents or due to better recollection of early life events (Sou et al, 2006). There was, however, no evidence of differential reporting of sick days for the first born children in our data as there was no statistically significant interaction between birth order and number of sick days.

Regarding social contacts, the above-mentioned diagnostic delay (Raaschou-Nielsen et al, 2006; ng-Tan and Franco, 2007, 2008; Wilne et al, 2007; Raab and Gartner, 2009) suggests confounding by indication as a potential concern as cases may have attended daycare less often compared with controls because of prodromal symptoms or other factors relating to a latent and undiagnosed CABT. This would lead to an underestimation of the association between social contacts and the risk of CABT.

Strengths of this study include the high participation rate, reducing the potential of selection bias, and the use of nationwide and high-quality registries (Tulinius et al, 1992; Storm et al, 1997; Barlow et al, 2009) in addition to case validation from treating physicians in hospitals ensuring a complete case ascertainment. Furthermore, cases were validated by unequivocal diagnostic imaging or histological confirmation ensuring correct diagnoses of included cases. The complete Scandinavian and Swiss population registers provided an optimal sampling frame for population-based controls.

Daycare attendance before the age of 6 years was high in all countries (range: 63–97%), making it a substantial source of social contacts with other children and therefore also for repeated contact with infectious agents. This in combination with a range of other measures of social contacts and infectious diseases allowed us to establish a quite detailed picture of the children’s total exposure. This included also social contacts at home, where we could account for the complex modern family structure with cohabiting full, half- and step-siblings, whereas previous studies have not had any such information or at best had registered maternal birth order (Emerson et al, 1991; Linet et al, 1996; Heuch et al, 1998; Mogren et al, 2003; Von Behren and Reynolds, 2003; Altieri et al, 2006; Shaw et al, 2006; Schmidt et al, 2010).

With such a generally high proportion of children attending daycare, it became difficult to investigate daycare as a risk factor; however, we created the score of childhours at daycare as exposure gradient and did not observe an association with CABT risk when comparing children of the highest and lowest quintile of childhours at daycare. The distribution of attending daycare was skewed between the countries; we did, however, not have sufficient statistical power to perform country-specific analyses, but sensitivity analysis, removing one country at a time, did not substantially alter the results (data not shown).

Whereas interview information about birth order and living with other children is expected to be accurate, information on daycare attendance may be difficult to recollect many years later. We believe, in contrast to the reported infectious diseases, that it is unlikely that case and control parents remember daycare attendance differently. There is, however, the possibility that non-differential bias could influence the risk estimates, possibly leading to a dilution of the risk.

In conclusion, this study provides little evidence of an association between increased social contacts in early life, as measured by daycare attendance, participation in baby groups or living together with other children, and the risk of CABT in 7–19 year olds. However, we observed more frequent reported sick days because of infections in the first 6 years of life associated with the risk of glioma and possibly embryonal brain tumour. Although the association could result from chance or recall bias, it needs to be further studied to identify whether infectious disease early in life could represent early symptoms of CABT, have the same underlying cause as CABT or modulate the risk of developing a CABT.