Overall, we found an inverse relationship between childhood cancer risk and increasing birth order. This effect was mainly seen in the CNS tumors, neuroblastoma, Wilms tumor, rhabdomyosarcoma, and bilateral retinoblastoma. Our pooled analysis included more than 17,000 children with cancer which allowed us to examine many rare sub-types with greater statistical power than in most previous studies. In addition, we were able to control for several factors, such as birth weight and maternal age that are associated with both birth order and several childhood cancers.
We observed only a slight, non-significant decrease in risk of ALL, the most common form of childhood leukemia, with increasing birth order. Previous studies of ALL and birth order have had inconsistent results with some studies reporting increased risk for ALL with high birth order 17, 18
, some finding decreased risk 7, 10, 11
, and others finding no association or weak associations 8, 9, 12
. In contrast to the ALL findings, we observed an increased risk with increasing birth order for the rarer types of leukemia, AML and CMD, primarily in the youngest age group. Several other studies that have examined AML separately have also found increased risk associated with increasing birth order, particularly in infants and young children 6-9, 19
although this was not reported by others 10-13
Our results for the CNS cancers are consistent with a recent case-control study from France that found decreased risk for 3rd
or higher birth order for CNS tumors (OR=0.8, 95% CI: 0.5-1.2) 20
. Significantly reduced risk for low grade gliomas and high birth order was also noted in a recent California study (25% of the cases in this new CA study were also included in the pooled dataset) 16
. Similarly Linet et al.
reported an increased risk for CNS tumors for first born children 21
. In contrast, Shaw et al.
reported that second or higher birth order children were at higher risk 22
. Several other studies have reported null or mixed and non-statistically significant results 23-26
. Two recent studies have suggested that CNS tumor risk increases with number of siblings 27
and children in the household 28
. These discrepancies may be due to relatively small sample sizes in many studies and different measures used, such as birth order vs. number of siblings.
Most previous studies of birth order and neuroblastoma have also shown that later born children are at decreased risk 29
. A recent French study found decreased neuroblastoma risk with high birth order with a very similar magnitude as that in our present study (OR=0.6, 95% CI 0.4, 1.0 for third or higher birth order) 30
Because of its rarity little is known about birth characteristics and Wilms tumors. Two previous studies of these embryonal tumors of the kidneys have reported, as did ours, that firstborn children are at increased risk, though neither of these smaller studies had statistically significant findings 31, 32
Similarly there is very little literature on retinoblastoma and birth order. A recent report from Australia found decreased risk for children who were not firstborn (OR= 0.86, 95% CI: 0.46, 1.64) 32
. It is not clear why the decreased risk we observed for increasing birth order would be for bilateral retinoblastoma only. This could be due to chance, given the large number of comparisons made and the relatively small number of retinoblastoma cases. Alternatively, families with bilateral retinoblastoma may receive genetic counseling about increased risks and opt to limit childbearing.
Birth order may be a marker of infectious exposures with later-born children presumed to be more often exposed by older siblings and exposed at earlier ages. The associations we observed between birth order and cancer risk for certain tumors suggests that the immune system may play some role in cancer risk. For example, Greaves proposed that delayed exposures to infections may cause an abnormal response after a common infection, increasing the chance of the second genetic mutation that leads to ALL 33
. However, we did not observe much difference in risk for ALL associated with birth order. Any relationship between birth order and infectious exposures may be diluted if the birth interval is large or if a child acquires infections from other sources, such as day care 34-36
. We were not able to account for either of these variables. We also lacked information on the number of household residents and other factors that affect children's immune systems, such as breast feeding, history of infectious illnesses, and vaccinations. Methods for improving the assessment of childhood infections are being developed, including use of clinical diagnoses via medical records 37
and daily diaries to capture sub-clinical infections, however the latter would be useful only in prospective studies 1
Birth order may be a marker of different hormonal exposures to the fetus. These early exposures to hormones may affect future cancer development 2
. It is possible that firstborn children have higher estrogen exposures that may contribute to greater risk of cancer than later born children. Estrogen levels in maternal and umbilical cord blood samples are somewhat greater in first pregnancies compared to second or third pregnancies 3, 4, 38
. Birth order also has been investigated with respect to several types of adult cancers, particularly those with possible hormonal-mediated mechanisms. Higher birth order has been associated with decreased risk of testicular cancer 39, 40
and adult glioma 41
. Many studies have examined birth order and breast cancer risk with mostly null findings 42
A third possibility is that higher birth order individuals have higher levels than first born individuals of microchimerism (presence of cells or DNA from genetically distinct individuals, in this instance as acquired in utero
from the mother and older siblings who may have exchanged cells with her during the earlier pregnancies). Bi-directional trafficking of cells between mother and fetus during pregnancy has been demonstrated 43
with retention of maternal cells among offspring potentially lasting decades 44
. Such microchimerism may result in varying levels of disease susceptibility by birth order 45
, and is one suggested mechanism 46
for an observed birth order pattern in the inheritance of chronic lymphocytic leukemia and lymphoproliferative disease 47
. However, this is purely speculation, as currently there is a paucity of epidemiologic data to explore this hypothesis.
This study was limited to information collected on birth certificates. We did not have a way to assess the accuracy of the birth order data. We also were not able to control for socioeconomic status, which may be associated with both birth order and some cancers 5
although SES, as measured by years of parental education, was not associated with cancer risk for most tumors types in this dataset 48
. In addition, we may have missed some cancer cases among the control subjects that could have moved out of the registry catchment areas.
In conclusion, this study had the advantage of a very large sample size drawn from population-based birth and cancer registries from five different states. We had information on many potential confounders such as maternal age, race, and birth weight. While we may have accurately identified associations of particular childhood cancers with birth order, the biologic mechanisms remain to be elucidated.