1. ¹Department of Epidemiology Research, Statens Serum Institut, Copenhagen, Denmark
2. ²School of Population Health, University of Auckland, Auckland, New Zealand
3. ³Bandim Health Project, Statens Serum Institut, Copenhagen, Denmark
4. ⁴Pediatric Clinic, The University Hospital Rigshospitalet and The Medical Faculty, University of Copenhagen, Copenhagen, Denmark

Correspondence: M Kamper-Jørgensen, E-mail: mka@ssi.dk

⁵The author holds the Danish Childhood Cancer Foundation professorship in Pediatric Oncology.

Based on the two-hit hypothesis, exposure to infections early in life has been hypothesized to reduce the risk of childhood acute lymphoblastic leukemia (ALL).¹ The first hit should occur prenatally, whereas the second hit is related to delayed or diminished exposure to infections in early childhood, accompanied by an abnormal immune response.^{1, 2} So far, no specific infectious agents affecting the risk of childhood ALL have been identified. Instead, established risk factors for infections early in life, such as attendance to childcare, have been commonly used as a proxy measure of an early high infection load. Previous studies on childcare attendance have presented conflicting results, which could reflect childcare facility differences or that the effect of childcare on childhood ALL is modified by other factors. To evaluate the association between childcare attendance and childhood ALL, we conducted a nationwide study using Danish register data. The study was carried out as a matched case–control study nested in the cohort of all Danish children in the period 1989–2004. All incident cases of ALL diagnosed in children aged 0–15 years in Denmark were identified. Ten population controls per case were individually matched on date of birth and sex. Data were retrieved from three population-based registries: The Nordic Society of Paediatric Haematology and Oncology (NOPHO) database,³ the Danish Civil Registration System⁴ and the Childcare Database.⁵

During the study period, 559 children were diagnosed with childhood ALL in Denmark. The immune phenotype of 467 cases (84%) was B-cell precursor ALL, 73 cases (13%) had T-lineage ALL, 7 cases (1%) had mature B-cell ALL and for 12 cases (2%) information about immune phenotype was lacking in the register. The leukemic karyotype classified 199 cases (36%) as common ALL (t(12;21)(q13;q22) translocation or a hyperdiploid karyotype), whereas the karyotype was lacking or consisted of other aberrations for the remaining 360 cases. A white blood cell count of <50 was found in 430 cases (77%), whereas 118 cases (21%) had a white blood cell count of 50. White blood cell count was not available for 11 cases. Median age at diagnosis was 4.0 years. Ten cases (2%) were diagnosed with childhood ALL before the age of 1.0 year. In Table 1, characteristics of cases and controls are shown. Information on childcare attendance during the first 2 years of life was complete for 176 cases (32%) and 1571 controls (28%). Overall, ever attendance to childcare was associated with a rate ratio (RR) of 0.68 (95% CI 0.48; 0.95). Children with incomplete registration of childcare but with one or more registration of childcare had a RR of 0.59 (0.43; 0.81). Each additional month enrolled in childcare was associated with an insignificantly decreasing RR of 0.98 (95% CI 0.94; 1.03). There was no statistically significant difference between different types of childcare (P=0.57). Increased risks of childhood ALL, however not significant, were seen with increasing birth weight and increasing maternal and paternal age. No significant main effect of birth order or population density on ALL risk was observed. Significantly fewer cases than controls were born outside of Denmark (4 vs 10%, P<0.01). None of the covariates listed in Table 1 changed the estimated effect of ever attendance to childcare on ALL risk by 5% or more; thus, crude estimates are presented. The variable changing the estimated effect of ever attendance to childcare the most was birth weight. When adjusted for birth weight, the estimate for ever attendance to childcare decreased to 0.65 (95% CI 0.47; 0.92).

Table 1 - Cases, controls and RR of childhood ALL according to childcare characteristics and covariates.

Full table (162K)

In the analyses presented in Tables 2, 3 and 4, a total of 142 cases and 1545 controls were included in the 'Ever attendance' category. Besides the 110 cases and 1114 controls with complete registration of childcare who ever attended childcare, the category included an additional 32 cases and 431 controls with one or more enrollments during follow-up, but with an incomplete registration of childcare attendance. The distribution of clinical characteristics did not differ between children ever attending childcare, never attending childcare and children with incomplete registration. Table 2 presents cases, matched controls and RR (95% CI) of childhood ALL according to attendance to childcare and childhood ALL subtype. Ever attendance to childcare was associated with a similarly reduced risk of all childhood ALL subtypes, albeit not significantly for subtypes with few cases. Since the risk reduction was comparable for all subtypes, we analyzed all cases together in the subsequent analyses. Table 3 shows cases, matched controls and RR (95% CI) of childhood ALL according to age at exposure to childcare. Regardless of the age at exposure, ever attendance to childcare was found to impose a decreased risk of childhood ALL. A test for homogeneity showed that the lower risk could not be attributed to any special ages at exposure. We also evaluated the role of exposure at different ages by studying the effect of age at enrollment (data not shown). Since we stopped follow-up at 2 years of age, and very few children in that age group are withdrawn from childcare once enrolled, the effect of age at enrollment was almost identical to the effect of time enrolled in childcare presented in Table 1. Finally, as a sensitivity analysis, we studied different time windows of ever attendance to childcare by varying the definition of exposure as <6, <12, <18 and <24 months of age. The different measures led to similar results (data not shown). Table 4 presents interaction analyses between attendance to childcare and covariates. Girls attending childcare had a statistically significantly decreased risk of ALL of 53% (95% CI 25%; 71%), whereas the risk was not significantly reduced for boys. However, the test for interaction revealed that the effect of childcare attendance was not significantly different for girls and boys (P_inter=0.13). Although children with 1 older siblings had a more reduced risk than children with no older siblings if ever attending childcare, the risk reduction was not significantly different (P_inter=0.14). No interaction was observed with birth weight (P_inter=0.76) or population density (P_inter=0.24). With regard to parental age, children of parents aged 25 years or more had a more pronounced reduced risk of childhood ALL if ever attending childcare than did children of parents younger than 25 years. The reduction was not significantly different for children of different maternal and paternal ages, however (P_inter=0.16 and P_inter=0.08, respectively).

Table 2 - Cases, matched controls and RR of childhood ALL, according to ALL subtype.

Full table

Table 3 - Cases, matched controls and RR of childhood ALL, according to age at exposure to childcare.

Full table

Table 4 - RR of childhood ALL associated with childcare attendance, according to characteristics of the child and the child's family.

Full table

In accordance with most, but not all, previous studies, we found childcare attendance early in life to reduce the risk of childhood ALL.⁶ The 'delayed infection hypothesis' suggested by Greaves¹ addressed causality for B-cell precursor ALL, not least for the incidence peak age group. However, as in previous studies of subtypes of childhood ALL, we found that a reduced risk when attending childcare early in life was not specific to this subset of ALL, but applied to several subtypes of childhood ALL. Not all cases were studied for the t(12:21)(q13;q22) translocation and many G-band karyotypings were unsuccessful. Thus, reliable conclusions regarding the effect of childcare attendance with respect to karyotypes cannot be drawn. The effect of childcare did not seem to be limited to a certain age at exposure within the first two years of life. We tried identifying vulnerable periods by different measures, which all gave similar results. We found a more reduced risk in girls than boys ever attending childcare. This sex difference may reflect the biology of ALL, since boys and girls differ both in their distribution of subsets of ALL and in the course of their disease. The reduced risk was not significantly different for boys and girls, however. Also, the reduced risk when attending childcare was more pronounced for children with older siblings and for children of parents aged 25 years or more at the time of childbirth than that for other children; however, it was not statistically significant. The underlying mechanism may be the same in the two groups. It remains to be explored whether high parental age reflects biological factors related to the type of ALL such as accumulation of chromosomal aberrations and mutations, or behavioral characteristics, such as frequency of social contacts. Mean maternal age at time of birth in Denmark was 31.0 years in 2004, and women with a long education tend to postpone childbirth compared with those with a short education. ALL may be initiated in utero and has previously been shown to be associated with birth weight.⁶ We found a slightly increased risk with increasing birth weight, in accordance with a Nordic population-based analysis of ALL.⁷ A significantly lower proportion of cases than controls were born outside Denmark. This is compatible with the lower registered incidence of childhood ALL in China, India and Africa, areas from which a considerable part of children living in Denmark, but born outside Denmark, originate. However, it is uncertain whether this reflects biological, socioeconomic or behavioral factors. The fact that more controls than cases were born outside Denmark also influenced the registration of birth weight, maternal age and paternal age. Since children born outside Denmark have no registration of birth weight and the children's biological parents do not always live in Denmark, fewer controls than cases had records on birth weight and parental age. Thus, the reduced risk associated with 'Not registered' birth weight and paternal age reflects that fewer controls than cases were born in Denmark. In the present study, we took advantage of very detailed information on childcare attendance, childhood leukemia and characteristics of the child and the child's family drawn from independent registers. Thereby, we minimized the concern in many previous studies that estimates were influenced by selection and information bias, particularly recall bias. Information from national registers furthermore ensured that children from all socioeconomic groups and geographical areas were included. Some children with childhood ALL may have had a pre-diagnostic phase with recurrent infections, bone pains and/or fatigue that caused the parents to refrain from enrolling their child in childcare. However, since a lower RR if ever attending childcare was seen in children diagnosed at the age of 5 years or later compared with those diagnosed before 5 years of age, we find confounding by indication unlikely. Information on childcare during the first 2 years of life was complete for 1747 (28%) of the children. More children from later birth cohorts had complete registration of childcare than did children from early birth cohorts. This was due to increasing registration of childcare in Denmark until the mid-1990s, after which childcare attendance was registered for approximately 90% of all children in Denmark. However, owing to the matched design, estimates were not affected by the increasing childcare registration. Registration of childcare was not associated with any of the presented covariates. Unlike in many other countries, childcare attendance is not strongly associated with socioeconomic factors in Denmark, as all municipalities have an obligation to provide childcare, most often from 6 months of age. Childcare attendance is publicly subsidized for all families, and families pay a fee corresponding to a maximum of 30% of the actual cost. Furthermore, the childcare fee can be reduced or waived for low-income families. Children not attending childcare during the first years of life are kept at home mainly by parents on parental leave. Owing to lack of availability of such data in the registries, we could not include information on breastfeeding, which has previously been shown to be associated with a reduced risk of childhood ALL in some, but not all, studies. By the age of 6 months, 57% of children in Denmark enrolled in childcare compared with 71% in homecare are exclusively breastfed.⁸ Thus the effect of childcare presented in the present study may even be underestimated.

In conclusion, childcare attendance during the first 2 years of life was associated with a reduced risk of childhood ALL, which was not restricted to specific ALL subtypes or ages at exposure. The present study supports the 'delayed infection hypothesis' and indicates that the studied child and family characteristics play a minor role in the association between childcare attendance early in life and childhood ALL.