Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Fetal growth and the risk of childhood CNS tumors and lymphomas
in Western Australia
Elizabeth Milne1*, Crystal L. Laurvick1, Eve Blair1, Nicholas de Klerk1, Adrian K. Charles2,3 and Carol Bower1
1
Telethon Institute for Child Health Research, Centre for Child Health Research,
The University of Western Australia, Perth, Western Australia
2
Princess Margaret Hospital and King Edward Memorial Hospital for Women, Subiaco, Perth, Western Australia
3
Schools of Paediatrics and Child Health, and Women and Infants Health, University of Western Australia
The etiology of childhood cancers is largely unknown, although sis).22,23 The mitotic properties of IGFs, coupled with their ability
the early age at diagnosis has led to particular interest in in utero to inhibit cell death, are thought to enhance tumor growth.24–27
and perinatal factors. Birth weight is the most frequently studied
perinatal factor in relation to risk of childhood cancers, and We recently reported a moderately strong positive association
results have been inconsistent. We investigated whether the risk of between the risk of ALL and proportion of optimal birth weight.28
CNS tumors and lymphomas in children was associated with three These findings are consistent with a biologically plausible mecha-
measures of the appropriateness of intra-uterine growth: propor- nism of accelerated growth associated with the growth-promoting
tion of optimal birth weight (POBW), birth length (POBL) and effects of IGFs in normal as well as cancer cells, and/or failure of
weight for length (POWFL). A cohort of 576,633 infants born in apoptosis of pre-leukemic cells initiated by an in utero ‘first hitÕ.29
Western Australia in 1980–2004 were followed from birth to diag- A study of fetal growth and risk of non-CNS solid tumors is cur-
nosis of a CNS tumor (n 5 183) or lymphoma (n 5 84) before age rently under way within our group. The aim of the present study
15, death, or December 31, 2005, and analyzed with Cox regres- was to investigate whether the risk of childhood CNS tumors or
sion. Overall, there was little evidence of any association between
fetal growth and risk of CNS tumors, although risk of ependy- lymphomas is associated with the appropriateness of fetal growth.
moma/choroid plexus tumors was positively associated with
POBL and negatively associated with POWFL. The risk of Hodg-
kin and Burkitt lymphoma increased with increasing fetal growth Material and methods
among boys only, whereas the increased risk observed with non-
Hodgkin lymphoma was only in girls. These associations between We used the Western Australian Data Linkage System30 to
fetal growth and disease risk were also observed among children obtain information on all births from the midwives’ notifications
not classified as high birth weight, suggesting that accelerated and birth registrations in Western Australia for the period 1980–
growth is more important than birth weight per se. Results were 2004 and cancer registrations for children diagnosed in 1980–
similar when cases were compared with their unaffected siblings, 2005 while aged 0–14 years. Linkage to the Birth Defects Regis-
suggesting that the increased growth associated with cancer risk try31 identified all children with a birth defect. The data were
was not general to the family. The associations we observed are organized into mother-child pairs with a unique encrypted identi-
consistent with causal pathways involving fetal growth factors. fier for each individual. The case/noncase status for each child
' 2008 Wiley-Liss, Inc.
was also determined. Siblings of cases were identified as those
individuals who had the same maternal identifier as a cancer case.
Key words: CNS tumors; lymphoma; childhood; cohort study; intra- The final linked data file was de-identified to protect patient confi-
uterine growth dentiality and was current as at May 2006.
There were 613,613 birth registrations between 1980 and 2004.
The exclusions applied to the data in defining the study cohort are
The etiology of childhood cancers is largely unknown, although depicted in Figure 1. There were 29,969 noncases and 101 cancer
the early age at diagnosis has led to a growing interest in both cases with birth defects, including Down syndrome, that were
in utero and perinatal factors. Birth weight is the perinatal factor excluded from the study. However, cancer cases with a birth
that has been most frequently studied in relation to childhood defect caused by the cancer, such as hydrocephalus caused by a
cancer risk, and it is one of the few for which positive associations congenital brain tumor, or whose sole birth defect was the cancer
have been reported. Positive associations with birth weight (n 5 46), were retained in the dataset. We excluded children with
have generally been reported for acute lymphoblastic leukemia other birth defects because of the potential for confounding of the
(ALL),1–9 while the findings for other cancer types have been con- relationship between cancer risk and POBW; for example, some
flicting. For example, studies of childhood lymphomas have vari- birth defects are known to be related to both birth weight and risk
ously reported associations with low birth weight,9,10 high birth of some cancers.32,33 The cohort of eligible subjects comprised
weight,4,11,12 or have shown no relationship.13,14 The variation in
the findings for any given childhood cancer may be due in part to
differences in study design, sources of data, case definitions,
method of categorization of birth weight, and whether adjustment Abbreviations: ALL, acute lymphoblastic leukemia; CNS, central nerv-
ous system; HR, hazard ratio; ICCC3, International Classification of Child-
is made for gestational age. The latter is required to differentiate hood Cancer 3rd Edition; IGF, insulin-like growth factor; IRR, incidence
between an association with birth weight per se and one with rate ratio; NHL, non-Hodgkin lymphoma; OR, odds ratio; PNET, primitive
intra-uterine growth. neuroectodermal tumour; POBL, proportion of optimal birth length;
The mechanism(s) underlying the observed association between POBW, proportion of optimal birth weight; POWFL, proportion of optimal
weight for length; RR, relative risk; SD, standard deviation.
birth weight and some childhood cancers has not been well eluci- Grant sponsor: National Health and Medical Research Council of Aus-
dated. Birth weight has been positively correlated with insulin-like tralia; Grant number: 404086. Grant sponsor: NHMRC; Grant numbers:
growth factor-I (IGF-I) and to a lesser extent with insulin-like 353628, 353514.
growth factor-II (IGF-II),15,16 which could be explained by their *Correspondence to: Telethon Institute for Child Health Research, PO
roles in regulating the normal growth and differentiation of cells Box 855, West Perth, Western Australia 6872, Australia.
Fax: 161-08-9489-7700. E-mail: lizm@ichr.uwa.edu.au
and tissues during fetal development.17 Some tumors have been Received 27 November 2007; Accepted after revision 22 January 2008
shown to produce IGF-I, IGF-II or their binding proteins, or to DOI 10.1002/ijc.23486
possess IGF receptors18–21; IGF-I, in particular, has been shown to Published online 15 April 2008 in Wiley InterScience (www.interscience.
inhibit the normal process of programmed cell death (apopto- wiley.com).
576,366 noncases, 183 cases of CNS tumors and 84 cases of lym- nature of any growth deviation more specifically, we also investi-
phoma. Children with other cancer diagnoses (n 5 614) were gated the comparable measure of the appropriateness of longitudi-
included in the noncase group, as they were considered to be at nal growth ‘proportion of optimal birth lengthÕ (POBL) (primarily
risk of the cancers under study. reflecting skeletal growth); and the similarly derived ‘proportion
The outcomes of interest in this study were CNS tumors or lym- of optimal weight for lengthÕ (POWFL), a measure of the appro-
phomas diagnosed before age 15 years. We used the International priateness of birth weight for length (primarily reflecting soft tis-
Classification of Childhood Cancer 3rd Edition (ICCC3) codes34 to sue growth). To assess the importance of each of these measures
determine the specific cancer groupings to be investigated; these of intra-uterine growth to the risk of each childhood cancer, they
were astrocytoma, medulloblastoma/primitive neuroectodermal tu- were examined separately in the analysis. z-scores of POBW,
mor (PNET), ependymoma/choroid plexus tumor, other gliomas, POBL and POWFL were calculated so that the estimated change
other intracranial/spinal tumors, Hodgkin lymphoma, non-Hodgkin in the risk of a specific cancer was per standard deviation (SD),
lymphoma (except Burkitt lymphoma) (referred to as ‘NHLÕ hereon allowing the regression estimates of each measure to be compared
in this paper) and Burkitt lymphoma. The ICCC3 categories are directly. The population means and SDs used in the calculation of
shown in Appendix 1. The following categories were not analyzed the z-score for each variable were those shown for the noncase
individually as there were too few cases: four intracranial/spinal group in Table I. Additional explanatory variables that were inves-
germ cell tumors, one unspecified CNS tumor, five miscellaneous tigated include sex and birth order (‘first bornÕ compared with
lymphoreticular neoplasms and one unspecified lymphoma. How- ‘subsequently bornÕ).
ever, the five cases from the first two categories were included in the Cox proportional hazard regression models were initially fitted
analysis of ‘All CNS tumors combinedÕ. Children were followed-up in STATA version 9.036 using the fractional polynomial procedure
from birth to their date of diagnosis, 15th birthday, date of death, or to determine the appropriate form of each continuous variable to
31 December 2005 – whichever occurred first. include in the models. In each case, a linear term was found to be
The main explanatory variable was ‘proportion of optimal birth the most appropriate. The PHREG procedure in SAS version 9.137
weightÕ (POBW) whose calculation is described in detail else- was used for Cox regression analyses to determine the association
where.35 Briefly, it is an estimate of the appropriateness of intra- between the explanatory variables and the risk of being diagnosed
uterine growth and is the ratio of the observed birth weight to the with each cancer. Where appropriate, we investigated possible
‘optimal birth weightÕ. Optimal birth weight is calculated from a effect modification by sex and birth order, and produced sex- and
regression equation including terms for gestational duration, birth order-specific estimates. The ‘discreteÕ option in the PHREG
maternal height, parity and infant sex, derived from a total popula- procedure was used to perform a conditional logistic regression of
tion of singleton births that excluded risk factors for intra-uterine cancer cases matched to their unaffected siblings. All confidence
growth restriction, including maternal smoking. To identify the intervals (CIs) presented are 95% CIs.
438 MILNE ET AL.
Results
97.6 (10.4)
96.9 (10.6)
96.0 (10.6)
97.1 (11.4)
96.3 (10.2)
100.1 (12.0)
93.7 (8.8)
97.1 (9.1)
99.8 (7.7)
97.5 (8.1)
A total of 183 CNS tumors and 84 lymphomas were identified.
Mean (SD)
POWFL
The numbers of each type are shown in Table I, along with other
Includes 18 Ependymomas and 3 choroid plexus tumors.–2Includes 32 Medulloblastomas and 6 PNETs.–3Includes 4 intracranial/spinal germ cell tumors and 1 unspecified CNS tumor.
descriptive characteristics. As previously indicated, some very
rare tumours—including 6 lymphomas—were not analysed indi-
vidually and are not presented in the tables.
The majority of CNS tumors and all but four of the lymphoma
cases were diagnosed at or after 2 years of age (data not shown).
99.5 (4.5)
99.6 (4.8)
101.0 (4.0)
99.3 (3.5)
98.5 (4.1)
100.2 (5.0)
99.4 (4.5)
101.3 (4.2)
100.0 (4.5)
98.7 (5.0)
Mean (SD)
Mean parental age was similar among the noncases and all cancer
POBL
SD, Standard deviation; POBW, proportion of optimal birth weight; POBL, proportion of optimal birth length; POWFL, proportion of optimal weight for length.
compared with noncases (data not shown).
The results of the univariate Cox regression analysis are shown
in Table II. The risk of CNS tumors was similar in boys and girls,
97.5 (12.5)
96.9 (13.5)
95.0 (10.7)
96.7 (10.6)
94.9 (12.3)
97.6 (12.4)
96.2 (12.5)
98.1 (11.9)
99.2 (13.7)
TABLE I – DESCRIPTIVE CHARACTERISTICS OF STUDY COHORT BY ICCC3 CNS TUMOUR AND LYMPHOMA CATEGORIES
101.4 (9.9)
while boys appeared to have a higher risk of Burkitt and Hodgkin
Mean (SD)
POBW
lymphomas than girls. There was some evidence that higher birth
order was associated with an increased risk of ‘Other gliomasÕ,
and a reduced risk of ‘Other intracranial/spinal tumorsÕ (Table II).
Overall, there was little evidence of any association between the
three measures of fetal growth and risk of CNS tumors, although
there was a suggestion that risk of ependymoma/choroid plexus
280,449 (48.7)
32 (40.5)
9 (42.9)
14 (63.6)
18 (47.4)
4 (22.2)
80 (43.7)
15 (50.0)
14 (50.0)
9 (45.0)
association between risk of NHL and POBW and POBL was only
Males N (%)
Non-Hodgkin lymphoma
Other intracranial/spinal
phomas (in boys) and NHL (in girls) who did not have ‘high birth
weightÕ according to two commonly used definitions: >3,500 g
Hodgkin lymphoma
Astrocytoma 1.07 0.69–1.66 0.80 0.51–1.25 0.95 0.76–1.19 0.99 0.79–1.23 0.95 0.76–1.18
Ependymoma/choroid plexus1 0.95 0.40–2.23 0.82 0.34–1.94 0.82 0.53–1.26 1.37 0.90–2.06 0.68 0.44–1.06
Other gliomas 1.04 0.45–2.40 2.05 0.86–4.90 0.94 0.62–1.43 0.93 0.62–1.39 0.97 0.63–1.47
Medulloblastoma/PNETs2 0.76 0.40–1.44 1.01 0.53–1.91 0.81 0.59–1.12 0.81 0.62–1.06 0.86 0.62–1.20
Other intracranial/spinal 0.83 0.33–2.11 0.34 0.11–1.04 1.01 0.63–1.59 1.12 0.71–1.76 0.97 0.61–1.54
All CNS tumors combined3 0.90 0.68–1.21 0.90 0.67–1.20 0.90 0.77–1.04 0.96 0.83–1.11 0.89 0.77–1.04
Hodgkin lymphoma 0.52 0.24–1.11 1.28 0.63–2.62 1.34 0.95–1.88 1.38 0.98–1.94 1.25 0.88–1.76
Non-Hodgkin lymphoma 0.78 0.37–1.65 1.18 0.56–2.48 1.05 0.73–1.51 1.09 0.76–1.57 1.00 0.69–1.45
Burkitt lymphoma 0.26 0.09–0.78 0.95 0.39–2.30 1.15 0.75–1.76 0.84 0.57–1.23 1.27 0.84–1.94
HR, Hazard Ratio; CI, 95% Confidence interval; POBW, proportion of optimal birth weight; POBL, proportion of optimal birth length; POWFL, proportion of optimal weight for length.
1
Includes 18 Ependymomas and 3 choroid plexus tumors.–2Includes 32 Medulloblastomas and 6 PNETs.–3Includes 4 intracranial/spinal germ cell tumors and 1 unspecified CNS tumor.
*Reference categories: male sex and first born.
Astrocytoma 0.94 (0.69–1.29) 0.96 (0.70–1.32) 0.94 1.12 (0.82–1.53) 0.88 (0.66–1.17) 0.25 0.87 (0.63–1.20) 1.02 (0.75–1.40) 0.48
Ependymoma/choroid plexus1 0.77 (0.42–1.40) 0.87 (0.47–1.64) 0.77 1.17 (0.65–2.12) 1.59 (0.90–2.80) 0.46 0.68 (0.37–1.25) 0.68 (0.37–1.28) 0.99
Other gliomas 1.07 (0.60–1.93) 0.82 (0.45–1.49) 0.53 1.18 (0.66–2.11) 0.78 (0.54–1.14) 0.25 1.03 (0.57–1.86) 0.91 (0.50–1.64) 0.77
Medulloblastoma/PNETs2 0.88 (0.57–1.34) 0.73 (0.44–1.20) 0.57 0.87 (0.58–1.31) 0.77 (0.58–1.03) 0.63 0.90 (0.59–1.38) 0.81 (0.48–1.34) 0.72
Other intracranial/spinal 1.37 (0.77–2.44) 0.66 (0.33–1.34) 0.12 1.43 (0.80–2.57) 0.83 (0.47–1.46) 0.19 1.27 (0.70–2.31) 0.68 (0.34–1.36) 0.18
All CNS tumors combined3 0.93 (0.77–1.14) 0.85 (0.69–1.06) 0.54 1.05 (0.86–1.27) 0.88 (0.72–1.07) 0.23 0.90 (0.74–1.10) 0.88 (0.71–1.10) 0.90
Hodgkin lymphoma 1.67 (1.14–2.43) 0.81 (0.43–1.51) 0.05 1.65 (1.11–2.45) 0.95 (0.53–1.71) 0.13 1.54 (1.03–2.30) 0.79 (0.42–1.47) 0.08
Non-Hodgkin lymphoma 0.76 (0.46–1.25) 1.53 (0.91–2.55) 0.05 0.76 (0.54–1.08) 1.85 (1.12–3.06) 0.006 0.82 (0.50–1.35) 1.28 (0.75–2.18) 0.23
Burkitt lymphoma 1.35 (0.85–2.14) 0.56 (0.21–1.50) 0.11 0.91 (0.56–1.46) 0.74 (0.49–1.10) 0.51 1.50 (0.95–2.35) 0.61 (0.23–1.64) 0.10
HR, Hazard Ratio; CI, 95% Confidence interval; POBW, proportion of optimal birth weight; POBL, proportion of optimal birth length; POWFL, proportion of optimal weight for length.
1
Includes 18 Ependymomas and 3 choroid plexus tumors.–2Includes 32 Medulloblastomas and 6 PNETs.–3Includes 4 intracranial/spinal germ cell tumors and 1 unspecified CNS tumor.
439
440 MILNE ET AL.
Interaction p-value
cancer group had at least one sibling (Table V). The matched sib-
ling analyses for ependymoma/choroid plexus tumor were indica-
0.77
0.79
0.88
0.41
0.78
0.82
0.20
0.87
0.10
HR, Hazard Ratio; CI, 95% Confidence interval; POBW, proportion of optimal birth weight; POBL, proportion of optimal birth length; POWFL, proportion of optimal weight for length.
tive of an increased risk with higher POBL (OR: 2.97 per 1 SD
Includes 18 Ependymomas and 3 choroid plexus tumors.–2Includes 32 Medulloblastomas and 6 PNETs.–3Includes 4 intracranial/spinal germ cell tumors and 1 unspecified CNS tumor.
increase) and a reduced risk with higher POWFL (OR: 0.59) (Ta-
ble VI). POBL appeared to be positively related to risk of ‘Other
intracranial/spinal tumorsÕ (OR: 1.87), and inversely related to risk
0.98 (0.69–1.39)
0.73 (0.37–1.42)
1.01 (0.59–1.70)
0.74 (0.46–1.20)
0.83 (0.31–2.26)
0.87 (0.70–1.09)
1.56 (0.99–2.46)
1.04 (0.61–1.75)
1.78 (1.06–2.98)
Subsequent HR (CI)
Discussion
This study provides some new insights into the relationship
between intra-uterine growth and specific childhood cancers by
Interaction p-value
weight and CNS tumors have been inconsistent, and effect esti-
mates have generally lacked precision because of small numbers
of cases. Von Behren and Reynolds reported an increased risk of
astrocytoma among girls (but not boys) with birth weight >4,000 g
0.98 (0.69–1.38)
0.89 (0.46–1.74)
1.02 (0.60–1.72)
0.68 (0.42–1.09)
0.62 (0.23–1.69)
0.85 (0.68–1.06)
1.54 (0.97–2.45)
1.11 (0.66–1.86)
1.49 (0.82–2.69)
1.6 (0.6–4.2) for birth weight <2,500 g, and 2.3 (0.9–6.1) for birth
All CNS tumors combined3
Medulloblastoma/PNETs2
Astrocytoma 65 (82.3) 118 96.9 (12.9) 94.8 (11.1) 99.6 (4.6) 98.7(4.5) 96.9 (10.4) 95.7 (9.1)
Ependymoma/choroid plexus1 17 (81.0) 33 93.7 (9.1) 93.5 (14.2) 100.6 (4.3) 97.9 (4.0) 92.7 (6.8) 94.9 (12.2)
Other gliomas 20 (90.9) 31 97.1 (11.1) 97.1 (11.3) 99.4 (3.6) 98.5 (6.6) 97.4 (9.5) 98.8 (15.3)
Medulloblastoma/PNETs2 31 (81.6) 62 94.3 (11.9) 97.5 (12.5) 98.1 (4.0) 99.7 (4.0) 95.8 (10.2) 97.6 (10.4)
Other intracranial/spinal 12 (66.7) 24 95.1 (13.6) 94.1 (8.0) 99.6 (4.7) 97.7 (3.8) 95.2 (13.1) 96.4 (6.9)
All CNS tumors combined3 149 (81.4) 279 95.5 (12.0) 95.4 (11.5) 99.2 (4.4) 98.7 (4.5) 95.9 (10.0) 96.4 (10.4)
Hodgkin lymphoma 25 (83.3) 42 101.4 (10.2) 96.2 (12.8) 101.1 (4.4) 99.7 (4.7) 99.9 (8.1) 96.1 (10.2)
Non-Hodgkin lymphoma 24 (85.7) 51 98.1 (12.3) 99.2 (11.3) 100.3 (4.4) 98.7 (6.9) 97.2 (8.5) 99.4 (8.9)
Burkitt lymphoma 19 (95.0) 32 99.9 (13.8) 98.2 (11.0) 98.9 (5.0) 99.5 (4.4) 100.4 (12.2) 98.3 (9.4)
1
Includes 14 Ependymomas and 3 choroid plexus tumors.–2Includes 25 Medulloblastomas and 6 PNETs.–3Includes 4 intracranial/spinal germ
cell tumours and 1 unspecified CNS tumor.
The first was that in utero exposure to excess glucose would tend produced conflicting results. The results of two studies by Schuz
to increase fetal weight, while also modifying the trans-placental and coworkers,9,10 a study by McKinney et al.45 and two studies
response to carcinogenic factors. The second was that augmented by Roman and coworkers43,46 were indicative of some increased
rates of cell division and growth may increase the sensitivity of risk of NHL with low birth weight (<2,500 g). None of these stud-
the fetus to carcinogenic agents in utero. Schuz and Forman sug- ies adjusted for gestational age. On the other hand, Yeazel et al.
gested that regulators of embryonic brain development, such as reported an OR of 1.5 (1.0–2.5) among children at least 2 years
the TWIST transcription factor, might remain up-regulated by old at diagnosis with birth weight >4,000 g, after adjustment for
IGF-I.9 Another possibility is that increased growth factors may gestational age.12 Petridou et al. reported a 42% (4–92%) increase
prevent apoptosis occurring in cells that have had a ‘first hitÕ, in risk of NHL for each 500 g increase in birth weight,
while also giving rise to larger babies. after adjustment for ‘socioeconomic, perinatal and sun exposure
We found that the risk of Hodgkin lymphoma was associated variablesÕ (p 1035).44
with higher than expected birth weight, birth length, and birth Burkitt lymphoma was more common in boys, and risk
weight for length, particularly in boys. These findings were repli- appeared to be somewhat elevated among boys who had higher
cated in the combined-sexes matched siblings analysis, suggesting than expected birth weight and weight for length. Similar findings
that the increased fetal growth associated with risk of Hodgkin were seen in the combined-sexes matched analysis of siblings,
lymphoma is an individual rather than a familial characteristic. suggesting that the increased fetal growth associated with risk of
The relationship between intra-uterine growth (POBW) was even Burkitt lymphoma is specific to the individual and not general to
stronger among boys who did not have ‘high birth weightÕ, despite the family. Similar to the findings for Hodgkin lymphoma in boys
the small number of cases in the analysis. This finding supports and NHL in girls, the fact that the risk of Burkitt lymphoma
the notion that risk of Hodgkin lymphoma is associated with tended to be elevated among boys who did not have ‘high birth
accelerated growth rather than high birth weight per se. To our weightÕ suggests that accelerated growth is the more important
knowledge, only three previous studies have examined birth factor. To our knowledge, only one study has reported an associa-
weight in relation to risk of Hodgkin lymphoma separately from tion between birth weight and Burkitt lymphoma specifically4;
other lymphomas,12,43,44 and none found an association. that study also reported an increased risk with high birth weight
Our results were consistent with the risk of NHL being (>3,500 g) (IRR 10.5:1.1–102.6), after adjustment for gestational
higher—among girls only—primarily with greater skeletal growth age.
but also with somewhat more soft tissue at birth than expected. Three other previous studies have combined all lymphomas into
However, compared with their siblings, NHL cases tended to have one group for analysis. Savitz and Ananth reported a positive asso-
lower birth weight for their length, though with small numbers the ciation with birth weight >4,000 g compared with 2,500–4,000 g,
confidence interval is wide. Together, these findings suggest that although there were only 26 cases in the analysis and confidence
NHL cases may belong to families with a tendency toward accel- intervals were wide.11 McKinney et al. presented weak evidence
erated fetal growth. Similar to the findings for Hodgkin lymphoma of an increased risk with low birth weight (<2,500 g),13 and
in boys, the fact that the association between fetal growth and risk Adami et al. reported no association.14 None of these three studies
of NHL tended to be elevated among girls who did not have ‘high adjusted for gestational age. No previous studies appear to have
birth weightÕ suggests that the pertinent risk factor is accelerated directly compared the birth weight or intra-uterine growth of
growth. Previous studies of NHL in relation to birth weight have lymphoma cases with that of their siblings.
442 MILNE ET AL.
As in other studies, our results were based on relatively small we investigated associations between the measures of fetal growth
numbers of cases, so many of our findings are inconclusive. We and the risk of specific cancer subtypes. Analyzing CNS tumors or
also acknowledge that there is heterogeneity within some of the lymphomas as a group (as in some previous studies) may mask
ICCC3 tumor categories, such as in the ‘other gliomaÕ category; important associations with specific subtypes.
however, there were insufficient cases to conduct some of the sub- Selection bias is unlikely to have affected our results. The study
group analyses that would have been of interest. In addition, we included the entire population of children born in Western Aus-
investigated at least three explanatory variables and their interac- tralia between 1980 and 2004, although some outward migration
tions with sex and birth order for each of several cancer groups, so will have occurred resulting in a loss to follow-up. However, there
that some of the observed associations may be due to chance is no evidence to indicate that outward migration of children is
because of multiple comparisons. However, we believe we have related to the appropriateness of intra-uterine growth, a diagnosis
been appropriately conservative in the description and interpreta- of childhood cancer, or both. Recall bias is not a threat to validity
tion of our findings. as information on risk factors was collected prior to diagnosis.
The use of administrative health data generally involves some To our knowledge, ours is the first population-based study to
level of compromise in terms of the variables available for inclu- have examined appropriateness of fetal growth in relation to the
sion in the analysis, as the data are often collected for purposes risk of childhood CNS tumors and lymphomas, and to have
other than research. For example, we would have liked to investi- directly compared fetal growth among cases and their siblings.
gate whether there was any association between head circumfer- Given the rarity of these childhood cancers, collaborative studies
ence at birth and risk of CNS tumors in childhood. Samuelsen involving data pooling are most likely to produce precise esti-
et al.47 reported an increased risk of pilocytic astrocytoma (RR mates of the associations with fetal growth. We recommend that
1.34:95% CI: 1.09–1.65) and ependymoma (RR 1.36:95% CI: future studies take account of gestational age (at least) in the anal-
0.92–2.01) per 1-cm increase in head circumference, after adjust- ysis of associations between fetal growth and cancer risk, and
ment for birth weight, gestational age and sex, among children compare cases with their siblings, as such an approach is
aged 0–15 years. Although we had data on head circumference for more likely to elucidate possible causal pathways. Further
children born from 1990 onwards, it was missing for 53% of CNS research into the underlying biological mechanisms suggested by
tumors over the entire time period. the observed associations for these childhood cancers is also of a
This study has some important strengths. The examination of high priority.
the risk of childhood cancer associated with the appropriateness of
intra-uterine growth—rather than birth weight per se—represents
an advance relative to previous studies, particularly on those that Acknowledgements
did not adjust for gestational age. Our approach also obviated the The authors acknowledge the assistance provided by Ms. Diana
need to assign an arbitrary cut-off for ‘high birth weightÕ. Further- Rosman and Ms. Carol Garfield at the Western Australian Data
more, we were able to investigate the risk of childhood cancers in Linkage Unit and Drs. Tim Threlfall and Judith Thompson at the
relation to three distinct aspects of intra-uterine growth: propor- Western Australian Cancer Registry. The authors also thank
tion of optimum weight, proportion of optimum length and the Ms. Margaret Wood for extracting the linked data files and provid-
proportion of optimum weight for length. To the extent possible, ing advice on the cleaning and preparation of the data for analysis.
References
1. Hjalgrim LL, Rostgaard K, Hjalgrim H, Westergaard T, Thomassen 12. Yeazel MW, Ross JA, Buckley JD, Woods WG, Ruccione K, Robison
H, Forestier E, Gustafsson G, Kristinsson J, Melbye M, Schmiegelow LL. High birth weight and risk of specific childhood cancers: a report
K. Birth weight and risk for childhood leukemia in Denmark, Sweden, from the Children’s Cancer Group. J Pediatr 1997;131:671–7.
Norway, and Iceland. J Natl Cancer Inst 2004;96:1549–56. 13. McKinney PA, Juszczak E, Findlay E, Smith K, Thomson CS. Pre-
2. Hjalgrim LL, Westergaard T, Rostgaard K, Schmiegelow K, Melbye and perinatal risk factors for childhood leukaemia and other malig-
M, Hjalgrim H, Engels EA. Birth weight as a risk factor for childhood nancies: a Scottish case control study. Br J Cancer 1999;80:1844–51.
leukemia: a meta-analysis of 18 epidemiologic studies. Am J Epide- 14. Adami J, Glimelius B, Cnattingius S, Ekbom A, Zahm SH, Linet M,
miol 2003;158:724–35. Zack M. Maternal and perinatal factors associated with non-Hodg-
3. Jourdan-Da Silva N, Perel Y, Mechinaud F, Plouvier E, Gandemer V, kin’s lymphoma among children. Int J Cancer 1996;65:774–7.
Lutz P, Vannier JP, Lamagnere JL, Margueritte G, Boutard P, Robert 15. Murphy VE, Smith R, Giles WB, Clifton VL. Endocrine regulation of
A, Armari C, et al. Infectious diseases in the first year of life, perinatal human fetal growth: the role of the mother, placenta, and fetus.
characteristics and childhood acute leukaemia. Br J Cancer 2004; Endocr Rev 2006;27:141–69.
90:139–45. 16. Ross JA, Perentesis JP, Robison LL, Davies SM. Big babies and
4. Lee J, Chia K-S, Cheung K-H, Chia S-E, Lee H-P. Birthweight and infant leukemia: a role for insulin-like growth factor-1? Cancer
the risk of early childhood cancer among Chinese in Singapore. Int J Causes Control 1996;7:553–9.
Cancer 2004;110:465–7. 17. LeRoith D, Adamo M, Werner H, Roberts C. Insulinlike growth fac-
5. Ma X, Metayer C, Does MB, Buffler PA. Maternal pregnancy loss, tors and their receptors as growth regulators in normal physiology and
birth characteristics, and childhood leukemia (United States). Cancer pathogenic states. Trends Endocrinol Metab 1991;2:134–9.
Causes Control 2005;16:1075–83. 18. Hirschfeld S, Helman L. Diverse roles of insulin-like growth factors
6. McLaughlin CC, Baptiste MS, Schymura MJ, Nasca PC, Zdeb MS. in pediatric solid tumors. In Vivo 1994;8:81–90.
Birth weight, maternal weight and childhood leukaemia. Br J Cancer 19. Campbell PG, Novak JF. Insulin-like growth factor binding protein
2006;94:1738–44. (IGFBP) inhibits IGF action on human osteosarcoma cells. J Cell
7. Paltiel O, Harlap S, Deutsch L, Knaanie A, Massalha S, Tiram E, Physiol 1991;149:293–300.
Barchana M, Friedlander Y. Birth weight and other risk factors for 20. Antoniades HN, Galanopoulos T, Neville-Golden J, Maxwell M.
acute leukemia in the Jerusalem Perinatal Study cohort. Cancer Epide- Expression of insulin-like growth factors I and II and their receptor
miol Biomarkers Preview 2004;13:1057–64. mRNAs in primary human astrocytomas and meningiomas; in vivo
8. Podvin D, Kuehn CM, Mueller BA, Williams M. Maternal and birth studies using in situ hybridization and immunocytochemistry. Int J
characteristics in relation to childhood leukaemia. Paediatr Perinat Cancer 1992;50:215–22.
Epidemiol 2006;20:312–22. 21. Boulle N, Logie A, Gicquel C, Perin L, Le Bouc Y. Increased levels
9. Schuz J, Forman MR. Birthweight by gestational age and childhood of insulin-like growth factor II (IGF-II) and IGF-binding protein-2 are
cancer. Cancer Causes Control 2007;18:655–63. associated with malignancy in sporadic adrenocortical tumors. J Clin
10. Schuz J, Kaatsch P, Kaletsch U, Meinert R, Michaelis J. Association Endocrinol Metab 1998;83:1713–20.
of childhood cancer with factors related to pregnancy and birth. Int J 22. Barres BA, Hart IK, Coles HS, Burne JF, Voyvodic JT, Richardson
Epidemiol 1999;28:631–9. WD, Raff MC. Cell death and control of cell survival in the oligoden-
11. Savitz DA, Ananth CV. Birth characteristics of childhood cancer drocyte lineage. Cell 1992;70:31–46.
cases, controls, and their siblings. Pediatr Hematol Oncol 1994;11: 23. Baserga R, Resnicoff M, D’Ambrosio C, Valentinis B. The role of the
587–99. IGF-I receptor in apoptosis. Vitam Horm 1997;53:65–98.
RISK OF CHILDHOOD CNS TUMORS AND LYMPHOMAS 443
24. Pollak MN, Schernhammer ES, Hankinson SE. Insulin-like growth 37. SAS Institute Inc. SAS v. 9.1, edn. 9.1 Cary North Carolina: SAS
factors and neoplasia. Nat Rev Cancer 2004;4:505–18. Institute Inc., 2003.
25. Werner H, Le Roith D. New concepts in regulation and function of 38. Von Behren J, Reynolds P. Birth characteristics and brain cancers in
the insulin-like growth factors: implications for understanding normal young children. Int J Epidemiol 2003;32:248–56.
growth and neoplasia. Cell Mol Life Sci 2000;57:932–42. 39. Emerson JC, Malone KE, Daling JR, Starzyk P. Childhood brain tu-
26. Baserga R, Resnicoff M, Dews M. The IGF-I receptor and cancer. En- mor risk in relation to birth characteristics. J Clin Epidemiol
docrine 1997;7:99–102. 1991;44:1159–66.
27. Bentov I, Werner H. IGF. IGF receptor and overgrowth syndromes. 40. Kuijten RR, Bunin GR, Nass CC, Meadows AT. Gestational and fam-
Pediatr Endocrinol Rev 2004;1:352–60. ilial risk factors for childhood astrocytoma: results of a case-control
28. Milne E, Laurvick CL, Blair E, Bower C, de Klerk N. Fetal growth study. Cancer Res 1990;50:2608–12.
and acute childhood leukemia: looking beyond birth weight. Am J 41. Mogren I, Malmer B, Tavelin B, Damber L. Reproductive factors
Epidemiol 2007;166:151–9. have low impact on the risk of different primary brain tumours in off-
29. Greaves M. Molecular genetics, natural history and the demise of spring. Neuroepidemiology 2003;22:249–54.
childhood leukaemia. Eur J Cancer 1999;35:1941–53. 42. Schuz J, Kaletsch U, Kaatsch P, Meinert R, Michaelis J. Risk factors
30. Holman CD, Bass AJ, Rouse IL, Hobbs MS. Population-based linkage for pediatric tumors of the central nervous system: results from a Ger-
of health records in Western Australia: development of a health serv- man population-based case-control study. Med Pediatric Oncol
ices research linked database.[see comment]. Aust N Z J Public 2001;36:274–82.
Health 1999;23:453–9. 43. Roman E, Simpson J, Ansell P, Lightfoot T, Mitchell C, Eden TO.
31. Bower C, Rudy E, Ryan A, Cosgrove P, Callaghan A. Report of the Perinatal and reproductive factors: a report on haematological malig-
birth defects registry of western Australia 1980–2005. No. 13. King nancies from the UKCCS. Eur J Cancer 2005;41:749–59.
Edward Memorial Hospital, Centre for Women’s Health, 2006. 44. Petridou ET, Dikalioti SK, Skalkidou A, Andrie E, Dessypris N, Tri-
32. Olshan AF. Wilms’ tumor, overgrowth, and fetal growth factors: a hy- chopoulos D. Sun exposure, birth weight, and childhood lymphomas:
pothesis. Cancer Genet Cytogenet 1986;21:303–7. a case control study in Greece. Cancer Causes Control 2007;18:1031–
33. Gracia Bouthelier R, Lapunzina P. Follow-up and risk of tumors in 7.
overgrowth syndromes. J Pediatr Endocrinol Metab 2005;18 (Suppl 45. McKinney PA, Cartwright RA, Saiu JM, Mann JR, Stiller CA, Draper
1):1227–35. GJ, Hartley AL, Hopton PA, Birch JM, Waterhouse JA, Johnston HE.
34. Steliarova-Foucher E, Stiller C, Lacour B, Kaatsch P. International The inter-regional epidemiological study of childhood cancer
classification of childhood cancer, third edition. Cancer 2005;103: (IRESCC): a case control study of aetiological factors in leukaemia
1457–67. and lymphoma. Arch Dis Child 1987;62:279–87.
35. Blair E, Liu X, de Klerk N, Lawrence D. Optimal fetal growth for the 46. Roman E, Ansell P, Bull D. Leukaemia and non-Hodgkin’s lym-
Caucasian singleton and assessment of appropriateness of fetal phoma in children and young adults: are prenatal and neonatal factors
growth: an analysis of a total population perinatal database. BMC important determinants of disease? Br J Cancer 1997;76:406–15.
Paediatrics 2005;5:13. 47. Samuelsen SO, Bakketeig LS, Tretli S, Johannesen TB, Magnus P.
36. StataCorp. Stata Statistical Software: Release 9, edn. 9 College Sta- Head circumference at birth and risk of brain cancer in childhood: a
tion, Texas: StataCorp LP, 2005. population-based study. Lancet Oncol 2006;7:39–42.