Cancer Risk Among Children With Very Low Birth Weights Logan G. Spector, Susan E. Puumala, Susan E.

Carozza, Eric J. Chow, Erin E. Fox, Scott Horel, Kimberly J. Johnson, Colleen C. McLaughlin, Peggy Reynolds, Julie Von Behren and Beth A. Mueller Pediatrics 2009;124;96 DOI: 10.1542/peds.2008-3069

The online version of this article, along with updated information and services, is located on the World Wide Web at:
http://pediatrics.aappublications.org/content/124/1/96.full.html

PEDIATRICS is the official journal of the American Academy of Pediatrics. A monthly publication, it has been published continuously since 1948. PEDIATRICS is owned, published, and trademarked by the American Academy of Pediatrics, 141 Northwest Point Boulevard, Elk Grove Village, Illinois, 60007. Copyright 2009 by the American Academy of Pediatrics. All rights reserved. Print ISSN: 0031-4005. Online ISSN: 1098-4275.

Downloaded from pediatrics.aappublications.org at Indonesia:AAP Sponsored on December 5, 2011

Cancer Risk Among Children With Very Low Birth Weights

WHATS KNOWN ON THIS SUBJECT: There is a strong association of hepatoblastoma with VLBW. No studies have examined VLBW and other childhood cancers adequately. WHAT THIS STUDY ADDS: This study examined VLBW and childhood cancer in a large, population-based study with 17 000 case subjects and 56 000 control subjects.
CONTRIBUTORS: Logan G. Spector, PhD,a,b Susan E. Puumala, MS,a Susan E. Carozza, PhD,c Eric J. Chow, MD, MPH,d Erin E. Fox, PhD,e Scott Horel, MS,c Kimberly J. Johnson, PhD,a Colleen C. McLaughlin, PhD,f Peggy Reynolds, PhD,g Julie Von Behren, MPH,g and Beth A. Mueller, DrPHd
aDivision of Epidemiology/Clinical Research, Department of Pediatrics and bMasonic Cancer Center, University of Minnesota, Minneapolis, Minnesota; cDepartment of Epidemiology and Biostatistics, School of Rural Public Health, Texas A & M Health Sciences Center, College Station, Texas; dDepartment of Pediatric Hematology-Oncology, Fred Hutchinson Cancer Research Center, Seattle, Washington; eCancer Epidemiology and Surveillance Branch, Texas Department of State Health Services, Austin, Texas; fNew York State Cancer Registry, New York Department of Health, Albany, New York; gNorthern California Cancer Center, Union City, California

abstract
OBJECTIVE: The risk of hepatoblastoma is strongly increased among children with very low birth weight (1500 g). Because data on very low birth weight and other childhood cancers are sparse, we examined the risk of malignancy with very low birth weight in a large data set. METHODS: We combined case-control data sets created by linking the cancer and birth registries of California, Minnesota, New York, Texas, and Washington states, which included 17 672 children diagnosed as having cancer at 0 to 14 years of age and 57 966 randomly selected control subjects. Unconditional logistic regression analysis was used to examine the association of cancer with very low birth weight and moderately low birth weight (1500 1999 g and 2000 2499 g, respectively), compared with moderate/high birth weight (2500 g), with adjustment for gender, gestational age, birth order, plurality, maternal age, maternal race, state, and year of birth. RESULTS: Most childhood cancers were not associated with low birth weights. However, retinoblastomas and gliomas other than astrocytomas and ependymomas were possibly associated with very low birth weight. The risk of other gliomas was also increased among children weighing 1500 to 1999 g at birth. CONCLUSIONS: These data suggested no association between most cancers and very low birth weight, with the exception of the known association of hepatoblastoma and possibly moderately increased risks of other gliomas and retinoblastoma, which may warrant conrmation. Pediatrics 2009;124:96104

KEY WORDS infant, very low birth weight, cancer, case-control studies, registries ABBREVIATIONS CI condence interval OR odds ratio VLBWvery low birth weight www.pediatrics.org/cgi/doi/10.1542/peds.2008-3069 doi:10.1542/peds.2008-3069 Accepted for publication Jan 7, 2009 Address correspondence to Logan G. Spector, PhD, Division of Epidemiology/Clinical Research, Department of Pediatrics, University of Minnesota, 420 Delaware St, SE, MMC 715, Minneapolis, MN 55455. E-mail: spector@umn.edu PEDIATRICS (ISSN Numbers: Print, 0031-4005; Online, 1098-4275). Copyright 2009 by the American Academy of Pediatrics FINANCIAL DISCLOSURE: The authors have indicated they have no nancial relationships relevant to this article to disclose.

96

SPECTOR et al

Downloaded from pediatrics.aappublications.org at Indonesia:AAP Sponsored on December 5, 2011

ARTICLES

Because survival rates for infants of very low birth weight (VLBW) (usually dened as 1500 g) have improved markedly in recent decades,1 the longterm health of these children requires increasing attention, with much research focusing on neurocognitive outcomes,2,3 growth and development,4,5 and the metabolic syndrome.6,7 Childhood cancer has not been studied often, but it was demonstrated that the incidence of a rare embryonal liver tumor, hepatoblastoma, is 1 order of magnitude greater among VLBW children than among children with moderate birth weights.812 Because VLBW infants are exposed to multiple medical interventions in NICUs13,14 at a time in development when antioxidant capacity is decreased15,16 and xenobioticmetabolizing enzyme expression is variable,17,18 an iatrogenic cause is plausible. Other childhood cancers also might be associated with VLBW, for similar reasons, although no similar-sized associations have been noted to date. Although birth weight has been examined frequently in relation to various childhood cancers,1921 minimally or moderately increased risks associated with VLBW might not have been detected. Nearly all studies grouped all infants of 2500 g, which possibly obscured differences in biological features and medical exposures between moderately low birth weight and VLBW infants. Moreover, because the prevalence of VLBW is low (presently 1% of births in the United States1), the large sample sizes required for detection of small/moderate relative risks gener-

ally are not available in studies of childhood cancers. Our objective was to ll a gap in the literature by examining whether specic childhood cancers, in addition to hepatoblastoma, demonstrate associations with VLBW. Although high birth weight is a risk factor for several pediatric malignancies,2023 we report here on VLBW alone, because the putative mechanism to explain increased cancer risk for large infants, namely, growth factor excess,22,24 differs from the mechanisms we hypothesize for small infants. Our analysis constitutes the rst comprehensive examination of cancer risk among children born at the lowest end of the birth weight distribution.

records, in ratios to case subjects ranging from 1:1 to 10:1, with frequency matching being used in 4 states and individual matching in 1 state (California). All states matched with respect to year of birth and 2 also matched with respect to gender.6,1619 Case subjects selected as control subjects in Minnesota and New York were excluded from analysis, as were subjects with reported Down syndrome. However, this condition was not recorded in Texas before 1984 and in Washington before 1989. Case subjects diagnosed before 28 days of life and control subjects who died before then also were excluded from analysis. The nal pooled data set consisted of 17 672 case subjects and 57 966 control subjects. Variable Specication Data on birth weight, gestational age, plurality, gender, birth order, year of birth, and maternal age and race were obtained from birth records. Maternal education was not recorded by all states until 1992 and so was used only in secondary analyses. Gestational ages calculated from last menstrual period dates and clinical estimates were provided by 4 states, whereas California recorded only the former. A combined gestational age variable was developed that gave preference to the calculated estimate when available and used the clinical estimate otherwise. Birth weights of 350 g and gestational lengths of 20 or 45 weeks were considered implausible and were treated as missing data. For sensitivity analysis, we also identied sub-

METHODS
Study Design We pooled data from populationbased, case-control studies conducted in California, Minnesota, New York (excluding New York City), Texas, and Washington states.8,10,12,25,26 Institutional review board approval for this analysis was obtained from each participating institution. Table 1 provides information about each component study. Cases of childhood cancer meeting eligibility criteria were identied in each population-based cancer registry and were matched to birth records by using sequential deterministic or probabilistic record linkage.27 Cases were classied according to the International Classication of Childhood Cancer, Third Edition.28 Control subjects were selected at random from birth

TABLE 1 Details of Studies Included in the Pooled Data Set

State California Minnesota New York Texas Washington Ages at Diagnoses 28 d to 4 y 28 d to 14 y 28 d to 14 y 28 d to 14 y 28 d to 14 y Years of Diagnoses 19881997 19882004 19852001 19901998 19802004 Years of Birth 19831997 19762004 19702001 19751998 19802004 No. of Case Subjects 4177 2170 4357 4647 2321 No. of Control Subjects 8730 8735 12 041 4732 23 728 Matching Factors Birth year, gender Birth year Birth year Birth year, gender Birth year

PEDIATRICS Volume 124, Number 1, July 2009

Downloaded from pediatrics.aappublications.org at Indonesia:AAP Sponsored on December 5, 2011

97

TABLE 2 Numbers of Cases According to Diagnosis

Cancer Type Leukemia Lymphoid leukemia Acute myeloid leukemia Chronic myeloproliferative diseases Lymphomas Hodgkins lymphoma Non-Hodgkins lymphoma Burkitts lymphoma Central nervous system tumors Ependymoma Astrocytoma Intracranial embryonal cell tumor Other gliomas Neuroblastoma Retinoblastoma Wilms tumor Hepatoblastoma Bone tumors Osteosarcoma Ewings sarcoma Soft-tissue sarcomas Rhabdomyosarcoma Fibrosarcoma Germ cell tumors Intracranial germ cell tumor Extracranial germ cell tumor Gonadal germ cell tumor Carcinomas Thyroid carcinoma Melanoma Total cancers Control No. of Casesa 5908 4756 849 110 1545 501 605 234 3807 390 1685 903 499 1482 684 1176 273 573 293 219 1067 583 131 572 131 134 281 409 159 126 17 672b 57 966

ables of state, year of delivery (1970 1985, 1986 1989, 1990 1993, or 1994 2004), and infant gender (matched in only 2 states). We decided a priori to adjust all risk estimates for maternal age (20, 20 24, 2529, 30 34, 3539, or 40 years), plurality (singleton or multiple birth), birth order (1, 2, 3, 4, or 5), and maternal race/ethnicity, because these variables are associated with birth weight33 and are established or suspected risk factors for several childhood cancers.34 Analyses of birth weight (1500, 1500 1999, 2000 2499, or 2500 g) were adjusted for gestational age (32, 3236, or 37 weeks) and vice versa. ORs with respect to birth weight and gestational age are presented for all cancers combined and for the 17 International Classication of Childhood Cancer, Third Edition cancer types with 200 cases in the data set (excluding heterogeneous other and unspecied tumor categories). Several secondary analyses were also conducted, including exclusion of subjects with implausible birth weights for gestational age, adjustment for maternal education for the years data were available, restriction to subjects with birth years coextensive with registry operation, and use of the alternate gestational age variable that gave preference to the clinical estimate.

Cases were grouped according to the International Classication of Childhood Cancer, Third Edition. a Cases without missing data for maternal race, maternal age, gender, gestational age, plurality, birth order, and birth weight. b The sum of specic diagnoses does not equal the total because unspecied tumors were omitted from the table.

jects with implausible birth weights for gestational age by using guidelines derived from expert clinical opinion.29 Continuous and discrete variables were grouped by using cutoff points determined a priori. Birth weight categories were 350 to 1499 g, 1500 to 1999 g, and 2000 to 2499 g, with a reference of 2500 g, which delineated groups with decreasing NICU utilization.30,31 Separating high (4000 g) from normal (2500 3999 g) birth weights and using the latter as reference only minimally affected results for low birth weights (data not shown). Therefore, high birth weights were included in the reference category of 2500 g, to simplify presentation. Gestational age categories were 32
98 SPECTOR et al

weeks, 32 to 36 weeks, and 37 weeks. Because methods for gestational age estimation changed over the long study period, which added to common concerns about the accuracy of early gestational ages,32 we did not pursue any ner discrimination of preterm weeks of gestation. Statistical Analyses Analyses were performed by using SAS 9.1 (SAS Institute, Cary, NC). Odds ratios (ORs) and 2-sided 95% condence intervals (CIs) were calculated by using unconditional logistic regression; individual matching in the California data set was broken to allow the use of this method. Risk estimates were adjusted for the matching vari-

RESULTS
Diagnoses were distributed approximately as would be expected on the basis of national surveillance data (Table 2).35 Male gender and white race were somewhat more frequent among case subjects, compared with control subjects; there were no appreciable differences regarding other variables (Table 3). Data for all variables (except maternal education) were available for 93.7% of case subjects and 92.7% of control subjects.

Downloaded from pediatrics.aappublications.org at Indonesia:AAP Sponsored on December 5, 2011

ARTICLES

TABLE 3 Characteristics of Patients With Childhood Cancer and Control Subjects

Variable Age at diagnosis 28 d to 4 y, n (%) 59 y, n (%) 1014 y, n (%) Mean SD, mo Maternal characteristics Age at delivery 20 y, n (%) 2024 y, n (%) 2529 y, n (%) 3034 y, n (%) 3539 y, n (%) 40 y, n (%) Mean SD, y Missing data, n Education 12 y, n (%) 12 y, n (%) 1316 y, n (%) 17 y, n (%) Mean SD, ya Missing data, n Race White, n (%) Black, n (%) Asian, n (%) Other, n (%) Missing data, n Childs characteristics Gender Male, n (%) Female, n (%) Missing data, n Birth weight 1500 g, n (%) 15001999 g, n (%) 20002499 g, n (%) 2500 g, n (%) Mean SD, g Missing data, n Gestational age 32 wk, n (%) 3236 wk, n (%) 37 wk, n (%) Mean SD, wk Missing data, n Plurality Singleton, n (%) Multiple, n (%) Missing data, n Birth order First, n (%) Second, n (%) Third, n (%) Fourth, n (%) Fifth or higher, n (%) Mean SD Missing data, n Case Subjects 10 977 (62.1) 3403 (19.3) 3292 (18.6) 62.42 50.80 Control Subjects

1821 (10.3) 4448 (25.2) 5682 (32.2) 3908 (22.1) 1537 (8.7) 270 (1.5) 26.91 5.70 6 2482 (20.2) 4346 (35.4) 4381 (35.7) 1053 (8.6) 12.66 2.88 5410 15 243 (87.8) 1147 (6.6) 694 (4.0) 284 (1.6) 304

6220 (10.7) 15 268 (26.3) 18 427 (31.8) 12 521 (21.6) 4670 (8.1) 839 (1.4) 26.715.67 21 7246 (19.2) 13 586 (36.1) 13 815 (36.7) 3017 (8.0) 12.71 2.79 20 302 48 568 (85.1) 3478 (6.1) 2590 (4.5) 2448 (4.3) 882

Childhood cancer overall and most cancer types, including those not shown, did not have increased ORs for associations with moderately low birth weight or VLBW (Table 4). Having a birth weight of 350 to 1499 g was associated with a greatly increased risk of hepatoblastoma (OR: 17.18 [95% CI: 7.46 39.54]), relative to weighing 2500 g at birth. Lessmarked but still increased ORs were observed for birth weights of 1500 to 1999 g (OR: 3.37 [95% CI: 1.44 7.88]) and 2000 to 2499 g (OR: 1.56 [95% CI: 0.812.98]). The large number of patients with hepatoblastomas with birth weights of 1500 g allowed further division of VLBW into categories of 350 to 749 g and 750 to 1499 g, with respective ORs of 46.78 (95% CI: 16.36 133.74) and 13.73 (95% CI: 5.6333.48). VLBW was associated with more than twofold increased ORs for 2 other tumor types, namely, gliomas other than astrocytomas and ependymomas (ie, other gliomas; birth weight of 1500 g, OR: 2.13 [95% CI: 0.71 6.39]; birth weight of 1500 1999 g, OR: 3.58 [95% CI: 1.98 6.47]) and retinoblastomas (birth weight of 1500 g, OR: 2.43 [95% CI: 1.00 5.89]). There was a signicant OR of 1.42 (95% CI: 1.011.99) for intracranial embryonal cell tumors associated with birth weights of 2000 to 2499 g alone. Exclusion of the 308 subjects with implausible birth weights for gestational age did not substantially alter most ORs or suggest associations with VLBW that were not apparent in the main analyses. However, the association of retinoblastoma with VLBW became markedly stronger (OR: 3.95 [95% CI: 1.5510.07]) and the OR for other gliomas increased slightly (OR: 2.60 [95% CI: 0.82 8.22]). Hepatoblastoma remained strongly associated with VLBW, but the OR was somewhat decreased (OR: 11.55 [95% CI: 4.5229.55]). Results of the other sensitivity analyses,
99

9728 (55.1) 7942 (44.9) 2 155 (0.9) 204 (1.2) 639 (3.7) 16 436 (94.3) 3420.66 592.64 238 236 (1.4) 1252 (7.4) 15 480 (91.2) 39.39 2.45 704 17 311 (98.0) 357 (2.0) 4 7186 (41.3) 5718 (32.9) 2745 (15.8) 1076 (6.2) 675 (3.9) 2.02 1.23 272

30 573 (52.8) 27 383 (47.2) 10 411 (0.7) 634 (1.1) 2108 (3.7) 54 438 (94.5) 3408.76 572.67 375 628 (1.1) 3776 (6.8) 50 858 (92.0) 39.52 2.40 2704 56 626 (97.7) 1307 (2.3) 33 23 131 (41.0) 18 383 (32.6) 9092 (16.1) 3454 (6.1) 2373 (4.2) 2.04 1.25 1533

a Years of education was not included as a continuous variable in New York from 1988 to 1990 because there were a greater number of missing data for the continuous variable; the categorical variable was used for comparison.

PEDIATRICS Volume 124, Number 1, July 2009

Downloaded from pediatrics.aappublications.org at Indonesia:AAP Sponsored on December 5, 2011

TABLE 4 Adjusted ORs and 95% CIs Relating Moderately Low Birth Weight and VLBW With Childhood
Cancers
Birth Weight 2500 g All cancers N OR (95% CI) Acute lymphoid leukemia N OR (95% CI) Acute myeloid leukemia N OR (95% CI) Hodgkins lymphoma N OR (95% CI) Non-Hodgkins lymphoma N OR (95% CI) Burkitts lymphoma N OR (95% CI) Ependymoma N OR (95% CI) Astrocytoma N OR (95% CI) Intracranial embryonal cell tumor N OR (95% CI) Other gliomas N OR (95% CI) Neuroblastoma N OR (95% CI) Retinoblastoma N OR (95% CI) Wilms tumor N OR (95% CI) Hepatoblastoma N OR (95% CI) Osteosarcoma N OR (95% CI) Ewings sarcoma N OR (95% CI) Rhabdomyosarcoma N OR (95% CI) Gonadal germ cell tumor N OR (95% CI) Control subjects, N 16 436 1 4492 1 791 1 450 1 552 1 224 1 373 1 1577 1 839 1 454 1 1386 1 637 1 1113 1 212 1 258 1 200 1 555 1 257 1 54 438 20002499 g 639 0.97 (0.871.08) 157 0.90 (0.751.08) 29 0.98 (0.651.47) 16 0.76 (0.421.37) 30 1.38 (0.902.12) 5 0.52 (0.201.32) 6 0.46 (0.201.05) 62 1.15 (0.871.52) 43 1.42 (1.011.99) 17 0.88 (0.511.51) 59 1.14 (0.851.52) 25 0.91 (0.591.41) 42 0.80 (0.571.13) 12 1.56 (0.812.98) 13 1.04 (0.551.97) 12 1.38 (0.722.64) 15 0.66 (0.381.13) 14 1.23 (0.662.31) 2108 15001999 g 204 0.98 (0.811.18) 44 0.81 (0.571.15) 9 1.06 (0.512.17) 6 1.23 (0.503.03) 6 1.00 (0.412.42) 2 5 1.19 (0.443.23) 10 0.57 (0.271.18) 7 0.68 (0.291.60) 17 3.58 (1.986.47) 23 1.35 (0.842.18) 8 0.97 (0.452.07) 14 0.77 (0.411.42) 8 3.37 (1.447.88) 1 3 5 0.77 (0.301.98) 2 634 3501499 g 155 1.15 (0.891.50) 22 0.65 (0.381.13) 8 1.39 (0.533.64) 2 4 0.85 (0.262.83) 1 3 11 1.48 (0.663.31) 5 1.06 (0.363.14) 5 2.13 (0.716.39) 13 1.03 (0.502.11) 12 2.43 (1.005.89) 7 0.83 (0.332.07) 41 17.18 (7.4639.54) 1 0 2 1 411

including adjustment of birth weight for maternal education and use of the alternate gestational age variable that gave preference to the clinical estimate, were not materially different from the results of the main analysis (data not shown). Although there were modestly increased ORs for several tumors in association with very preterm delivery (32 weeks), all CIs included 1 (Table 5). Wilms tumor (OR: 1.51 [95% CI: 1.211.88]) and Ewings sarcoma (OR: 1.68 [95% CI: 1.032.76]) were associated with moderately preterm delivery (3236 weeks). Use of the alternate gestational age variable that gave preference to the calculated estimate produced similar results (data not shown), with the exception that Ewings sarcoma displayed somewhat stronger ORs for moderately (OR: 2.14 [95% CI: 1.323.47]) and very (OR: 3.11 [95% CI: 0.89 10.88]) preterm delivery.

DISCUSSION
Because the risk of most cancers among children with the lowest birth weights has not been assessed adequately, we undertook to examine this issue with a large, population-based, data set. The strong inverse association between hepatoblastoma and birth weight was conrmed, with relative risks estimated more precisely here than in most previous studies.812 Most other cancers were not associated with moderately low birth weight or VLBW, although retinoblastoma and other gliomas displayed possibly increased risks. The size of association with VLBW that we were able to detect is an important consideration. Analysis of CIs is the preferred method for determining the size of association consistent with the data.36 Table 4 shows upper condence limits of 1.50 and 1.13 for all childhood cancers and acute lymphoid leukemia, respectively, whereas upper

N values indicate the numbers of case subjects without missing birth weight data. ORs were adjusted for gender, gestational age, plurality, birth order, maternal age, maternal race, state, and year of birth. indicates omission of data due to small cell counts.

100

SPECTOR et al

Downloaded from pediatrics.aappublications.org at Indonesia:AAP Sponsored on December 5, 2011

ARTICLES

TABLE 5 Adjusted ORs and 95% CIs Relating Gestational Age With Childhood Cancers
Gestational Age 37 wk All cancers N OR (95% CI) Acute lymphoid leukemia N OR (95% CI) Acute myeloid leukemia N OR (95% CI) Hodgkins lymphoma N OR (95% CI) Non-Hodgkins lymphoma N OR (95% CI) Burkitts lymphoma N OR (95% CI) Ependymoma N OR (95% CI) Astrocytoma N OR (95% CI) Intracranial embryonal cell tumor N OR (95% CI) Other gliomas N OR (95% CI) Neuroblastoma N OR (95% CI) Retinoblastoma N OR (95% CI) Wilms tumor N OR (95% CI) Hepatoblastoma N OR (95% CI) Osteosarcoma N OR (95% CI) Ewings sarcoma N OR (95% CI) Rhabdomyosarcoma N OR (95% CI) Gonadal germ cell tumor N OR (95% CI) Control subjects, N 15 480 1 4199 1 758 1 431 1 522 1 200 1 355 1 1495 1 803 1 424 1 1308 1 603 1 1023 1 207 1 251 1 181 1 522 1 249 1 50 858 3236 wk 1252 1.07 (0.991.15) 324 1.06 (0.931.21) 50 0.86 (0.631.17) 30 0.92 (0.611.40) 37 0.89 (0.611.28) 22 1.53 (0.942.47) 19 0.80 (0.491.31) 96 0.83 (0.661.05) 63 0.97 (0.731.29) 42 1.07 (0.741.56) 109 1.03 (0.831.29) 52 1.05 (0.771.45) 109 1.51 (1.211.88) 18 0.69 (0.391.23) 22 1.31 (0.812.11) 22 1.68 (1.032.76) 40 1.08 (0.761.53) 20 1.11 (0.681.81) 3776 32 wk 236 1.13 (0.921.40) 47 1.05 (0.721.54) 11 0.97 (0.452.12) 4 0.75 (0.242.38) 10 1.63 (0.733.63) 1 5 1.29 (0.423.95) 12 0.58 (0.281.20) 8 0.80 (0.341.90) 8 0.76 (0.311.88) 25 1.33 (0.782.29) 12 0.87 (0.382.03) 11 0.96 (0.462.03) 38 1.59 (0.703.64) 2 3 2.32 (0.678.02) 2 5 2.44 (0.896.74) 628

tions of VLBW with most types of childhood cancer were present in this data set, notwithstanding possible biases and other limitations to interpretation. The association of VLBW with hepatoblastoma is unlikely to be attributable to chance or bias, having been observed in the United Kingdom11 and Japan9 as well as in the current US study (which included data reported in 3 previous publications8,10,12). Our ndings regarding other gliomas and retinoblastomas might be attributable to random error, however, given the number of comparisons made; less weight also might be attributed to the observations, because there was a lack of clear evidence of a doseresponse relationship between lower birth weights and these cancers. If real, these associations warrant some explanation, although it would necessarily be speculative. We posited that the increased risk of hepatoblastoma among VLBW infants might be attributable to exposures in the NICU, and we sought to determine whether similar, but weaker, associations existed with other childhood cancers. The list of possible iatrogenic hazards in NICUs, as recently reviewed by Lai and Bearer,14 includes light, oxygen, irradiation, electromagnetic elds, plasticizers, medications, and total parenteral nutrition. Among these exposures, irradiation and oxidative stress (through high-fraction oxygenation or lipid peroxidation) stand out for their known carcinogenicity.37,38 The plasticizer di(2-ethylhexyl)phthalate is also considered a probable human carcinogen.39 The possible role of NICU exposures in causing hepatoblastoma was examined in 3 small Japanese studies with 5 to 15 case subjects; greater durations of oxygen therapy, furosemide use, and hospitalization were observed among VLBW case subjects, compared with weightmatched control subjects.4042 Larger
101

N values indicate the numbers of case subjects without missing birth weight data. ORs were adjusted for gender, birth weight, plurality, birth order, maternal age, maternal race, state, and year of birth.

condence limits ranging from 2 to 4 were found for most other cancers; ORs larger than the upper condence
PEDIATRICS Volume 124, Number 1, July 2009

limits are unlikely to have been missed. Therefore, our results indicate that no moderate to strong associa-

Downloaded from pediatrics.aappublications.org at Indonesia:AAP Sponsored on December 5, 2011

studies may provide more-denitive answers. One limitation of our record-linkage study is that we did not have access to NICU exposure information. The use of birth weight as a proxy for medical exposures also requires caution, for several reasons. First, even among VLBW infants, almost all of whom spend some time in a NICU, there are variations in treatment. For instance, among infants with birth weights of 501 to 1500 g who were tracked by the National Institute of Child Health and Human Development Neonatal Research Network in 1999 2000, 57% received surfactant treatment, 27% received oxygen therapy at 28 days of life, and 20% received postnatal steroid treatment.43 Moreover, because medical practices have changed over time (eg, no infants in the aforementioned series received surfactant in 1987 1988), VLBW might have acted as a proxy for differing exposures during the long period of this study. VLBW also may signal the underlying pathologic processes that produced small size, rather than (or in addition to) postnatal treatments, although the effect of such confounding is difcult to predict and may differ according to cancer type. Congenital anomalies are overrepresented both among VLBW infants, compared with infants with normal birth weights,44 and among children with cancer45; the net effect of these tendencies would be to exaggerate associations of VLBW with malignancies. Conversely, high birth weight was associated with risk of acute lymphoid leukemia, Wilms tumor, and astrocytoma in our study (data not shown), as in others,2023

which would tend to counter any increase in risk attributable to medical exposures among VLBW infants. In this context, we note that the OR of 1.48 relating VLBW to astrocytomas was possibly inconsistent with the overall signicant linear trend in the risk of this tumor with increasing birth weight. The ORs relating VLBW to acute lymphoid leukemia and Wilms tumor, in contrast, supported the overall linear trends. Without additional information about the causes of VLBW, our data could not address these concerns. We controlled for gestational age in analyses of birth weight and observed few associations of childhood cancers with preterm birth itself. However, we cannot dismiss the possibility that our results for VLBW were actually attributable to residual confounding by the length of gestation, because the limitations of measurement of gestational age in vital records necessitated the use of broad categories of preterm birth.32 The use of population-based cancer surveillance data and prospectively collected birth records were major strengths of this study. Birth weight in vital records is considered generally accurate, as are the other variables we used, apart from gestational age.46 Several limitations must be noted. We might have missed cancer diagnoses among control subjects who were sampled from birth years before the inception of their respective cancer registries or who moved out of state, but we estimate, on the basis of US surveillance data,35 that only 125 cancers would be expected among control subjects if all were lost to follow-up

monitoring. Therefore, even under the most extreme assumptions, disease misclassication would be unlikely to bias our results. Mortality rates among VLBW infants are also many times greater than those at higher birth weights.1 Limiting subjects to those who survived to 28 days of life minimized this issue, because most deaths among VLBW infants occur before that time.47 However, person-time at risk among VLBW control subjects was probably still overestimated, relative to children with normal birth weights, because later deaths were not accounted for in these data. This would result in underestimation of the ORs for VLBW, although it is not possible for us to gauge the extent of such bias. Lastly, we could not control completely for socioeconomic status, which is associated with both VLBW48 and some childhood cancers.49 However, the results of subanalyses adjusting for maternal education as a marker for socioeconomic status were concordant with those of the main analysis.

CONCLUSIONS
Medical exposures in the NICU, combined with the immature defenses of premature infants, may plausibly affect future cancer risk. Apart from hepatoblastoma, associations with most childhood cancers were not increased, although possible moderate increases in risks of other gliomas and retinoblastoma were noted and may warrant conrmation. These results may be reassuring to practitioners and families concerned with the longterm health of VLBW infants.

REFERENCES
1. Centers for Disease Control and Prevention. Infant mortality and low birth weight among black and white infants: United States, 1980 2000. MMWR Morb Mortal Wkly Rep. 2002;51(27):589 592 2. Moster D, Lie RT, Markestad T. Long-term medical and social consequences of preterm birth. N Engl J Med. 2008;359(3):262273

102

SPECTOR et al

Downloaded from pediatrics.aappublications.org at Indonesia:AAP Sponsored on December 5, 2011

ARTICLES

3. Schendel D, Bhasin TK. Birth weight and gestational age characteristics of children with autism, including a comparison with other developmental disabilities. Pediatrics. 2008;121(6):11551164 4. Saigal S, Stoskopf B, Streiner D, Paneth N, Pinelli J, Boyle M. Growth trajectories of extremely low birth weight infants from birth to young adulthood: a longitudinal, population-based study. Pediatr Res. 2006;60(6):751758 5. Iba n ez L, Ferrer A, Marcos MV, Hierro FR, de Zegher F. Early puberty: rapid progression and reduced nal height in girls with low birth weight. Pediatrics. 2000;106(5). Available at: www.pediatrics.org/cgi/content/full/106/5/e72 6. Finken MJ, Keijzer-Veen MG, Dekker FW, et al. Preterm birth and later insulin resistance: effects of birth weight and postnatal growth in a population based longitudinal study from birth into adult life. Diabetologia. 2006;49(3):478 485 7. Phillips DI. Fetal programming of the neuroendocrine response to stress: links between low birth weight and the metabolic syndrome. Endocr Res. 2004;30(4):819 826 8. Reynolds P, Urayama KY, Von Behren J, Feusner J. Birth characteristics and hepatoblastoma risk in young children. Cancer. 2004;100(5):1070 1076 9. Tanimura M, Matsui I, Abe J, et al. Increased risk of hepatoblastoma among immature children with a lower birth weight. Cancer Res. 1998;58(14):30323035 10. McLaughlin CC, Baptiste MS, Schymura MJ, Nasca PC, Zdeb MS. Maternal and infant birth characteristics and hepatoblastoma. Am J Epidemiol. 2006;163(9):818 828 11. Ansell P, Mitchell CD, Roman E, Simpson J, Birch JM, Eden TO. Relationships between perinatal and maternal characteristics and hepatoblastoma: a report from the UKCCS. Eur J Cancer. 2005;41(5): 741748 12. Spector LG, Johnson KJ, Soler JT, Puumala SE. Perinatal risk factors for hepatoblastoma. Br J Cancer. 2008;98(9):1570 1573 13. Clark RH, Bloom BT, Spitzer AR, Gerstmann DR. Reported medication use in the neonatal intensive care unit: data from a large national data set. Pediatrics. 2006;117(6):1979 1987 14. Lai TT, Bearer CF. Iatrogenic environmental hazards in the neonatal intensive care unit. Clin Perinatol. 2008;35(1):163181 15. Weinberger B, Laskin DL, Heck DE, Laskin JD. Oxygen toxicity in premature infants. Toxicol Appl Pharmacol. 2002;181(1):60 67 16. Dani C, Cecchi A, Bertini G. Role of oxidative stress as physiopathologic factor in the preterm infant. Minerva Pediatr. 2004;56(4):381394 17. Hines RN, McCarver DG. The ontogeny of human drug-metabolizing enzymes: phase I oxidative enzymes. J Pharmacol Exp Ther. 2002;300(2):355360 18. McCarver DG, Hines RN. The ontogeny of human drug-metabolizing enzymes: phase II conjugation enzymes and regulatory mechanisms. J Pharmacol Exp Ther. 2002;300(2):361366 19. Daling JR, Starzyk P, Olshan AF, Weiss NS. Birth weight and the incidence of childhood cancer. J Natl Cancer Inst. 1984;72(5):1039 1041 20. Harder T, Plagemann A, Harder A. Birth weight and subsequent risk of childhood primary brain tumors: a meta-analysis. Am J Epidemiol. 2008;168(4):366 373 21. Hjalgrim LL, Westergaard T, Rostgaard K, et al. Birth weight as a risk factor for childhood leukemia: a meta-analysis of 18 epidemiologic studies. Am J Epidemiol. 2003;158(8):724 735 22. Leisenring WM, Breslow NE, Evans IE, Beckwith JB, Coppes MJ, Grundy P. Increased birth weights of National Wilms Tumor Study patients suggest a growth factor excess. Cancer Res. 1994;54(17): 4680 4683 23. Olshan AF, Breslow NE, Falletta JM, et al. Risk factors for Wilms tumor: report from the National Wilms Tumor Study. Cancer. 1993;72(3):938 944 24. Ross JA, Perentesis JP, Robison LL, Davies SM. Big babies and infant leukemia: a role for insulinlike growth factor-1? Cancer Causes Control. 1996;7(5):553559 25. Chow EJ, Friedman DL, Mueller BA. Maternal and perinatal characteristics in relation to neuroblastoma. Cancer. 2007;109(5):983992 26. Walker KM, Carozza S, Cooper S, Elgethun K. Childhood cancer in Texas counties with moderate to intense agricultural activity. J Agric Saf Health. 2007;13(1):9 24 27. Jaro MA. Probabilistic linkage of large public health data les. Stat Med. 1995;14(57):491 498 28. Steliarova-Foucher E, Stiller C, Lacour B, Kaatsch P. International Classication of Childhood Cancer, Third Edition. Cancer. 2005;103(7):14571467 29. Alexander GR, Himes JH, Kaufman RB, Mor J, Kogan M. A United States national reference for fetal growth. Obstet Gynecol. 1996;87(2):163168 30. Russell RB, Green NS, Steiner CA, et al. Cost of hospitalization for preterm and low birth weight

PEDIATRICS Volume 124, Number 1, July 2009

Downloaded from pediatrics.aappublications.org at Indonesia:AAP Sponsored on December 5, 2011

103

31. 32. 33. 34.

35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47.

48.

49.

infants in the United States. Pediatrics. 2007;120(1). Available at: www.pediatrics.org/cgi/content/ full/120/1/e1 Schwartz RM, Kellogg R, Muri JH. Specialty newborn care: trends and issues. J Perinatol. 2000; 20(8):520 529 Wier ML, Pearl M, Kharrazi M. Gestational age estimation on United States livebirth certicates: a historical overview. Paediatr Perinat Epidemiol. 2007;21(suppl 2):4 12 Valero De Bernabe J, Soriano T, Albaladejo R, et al. Risk factors for low birth weight: a review. Eur J Obstet Gynecol Reprod Biol. 2004;116(1):315 Ries LA, Smith MA, Gurney JG, et al. Cancer Incidence and Survival Among Children and Adolescents: United States SEER Program 19751995. Bethesda, MD: National Cancer Institute; 1999 Ries LAG, Melbert D, Krapcho M, et al. SEER Cancer Statistics Review, 19752005. Bethesda, MD: National Cancer Institute; 2008 Smith AH, Bates MN. Condence limit analyses should replace power calculations in the interpretation of epidemiologic studies. Epidemiology. 1992;3(5):449 452 Caporaso N. The molecular epidemiology of oxidative damage to DNA and cancer. J Natl Cancer Inst. 2003;95(17):12631265 Wakeford R. The cancer epidemiology of radiation. Oncogene. 2004;23(38):6404 6428 Agency for Toxic Substances and Disease Registry. Toxicological Prole for Di(2-ethylhexyl)phthalate (DEHP). Atlanta, GA: Agency for Toxic Substances and Disease Registry; 2002 Maruyama K, Ikeda H, Koizumi T, Tsuchida Y. Prenatal and postnatal histories of very low birthweight infants who developed hepatoblastoma. Pediatr Int. 1999;41(1):82 89 Maruyama K, Ikeda H, Koizumi T, et al. Case-control study of perinatal factors and hepatoblastoma in children with an extremely low birthweight. Pediatr Int. 2000;42(5):492 498 Oue T, Kubota A, Okuyama H, et al. Hepatoblastoma in children of extremely low birth weight: a report from a single perinatal center. J Pediatr Surg. 2003;38(1):134 137 Fanaroff AA, Hack M, Walsh MC. The NICHD Neonatal Research Network: changes in practice and outcomes during the rst 15 years. Semin Perinatol. 2003;27(4):281287 Mili F, Edmonds LD, Khoury MJ, McClearn AB. Prevalence of birth defects among low-birth-weight infants: a population study. Am J Dis Child. 1991;145(11):13131318 Merks JH, Ozgen HM, Koster J, Zwinderman AH, Caron HN, Hennekam RC. Prevalence and patterns of morphological abnormalities in patients with childhood cancer. JAMA. 2008;299(1):61 69 Northam S, Knapp TR. The reliability and validity of birth certicates. J Obstet Gynecol Neonatal Nurs. 2006;35(1):312 Gould JB, Benitz WE, Liu H. Mortality and time to death in very low birth weight infants: California, 1987 and 1993. Pediatrics. 2000;105(3). Available at: www.pediatrics.org/cgi/content/full/105/3/ e37 Cubbin C, Marchi K, Lin M, et al. Is neighborhood deprivation independently associated with maternal and infant health? Evidence from Florida and Washington. Matern Child Health J. 2008; 12(1):6174 Little J. Epidemiology of Childhood Cancer. Lyon, France: International Agency for Research on Cancer; 1999

104

SPECTOR et al

Downloaded from pediatrics.aappublications.org at Indonesia:AAP Sponsored on December 5, 2011

Cancer Risk Among Children With Very Low Birth Weights Logan G. Spector, Susan E. Puumala, Susan E. Carozza, Eric J. Chow, Erin E. Fox, Scott Horel, Kimberly J. Johnson, Colleen C. McLaughlin, Peggy Reynolds, Julie Von Behren and Beth A. Mueller Pediatrics 2009;124;96 DOI: 10.1542/peds.2008-3069
Updated Information & Services References including high resolution figures, can be found at: http://pediatrics.aappublications.org/content/124/1/96.full.htm l This article cites 42 articles, 12 of which can be accessed free at: http://pediatrics.aappublications.org/content/124/1/96.full.htm l#ref-list-1 This article has been cited by 2 HighWire-hosted articles: http://pediatrics.aappublications.org/content/124/1/96.full.htm l#related-urls One P3R has been posted to this article: http://pediatrics.aappublications.org/cgi/eletters/124/1/96 This article, along with others on similar topics, appears in the following collection(s): Premature & Newborn http://pediatrics.aappublications.org/cgi/collection/premature _and_newborn Information about reproducing this article in parts (figures, tables) or in its entirety can be found online at: http://pediatrics.aappublications.org/site/misc/Permissions.xht ml Information about ordering reprints can be found online: http://pediatrics.aappublications.org/site/misc/reprints.xhtml

Citations

Post-Publication Peer Reviews (P3Rs) Subspecialty Collections

Permissions & Licensing

Reprints

PEDIATRICS is the official journal of the American Academy of Pediatrics. A monthly publication, it has been published continuously since 1948. PEDIATRICS is owned, published, and trademarked by the American Academy of Pediatrics, 141 Northwest Point Boulevard, Elk Grove Village, Illinois, 60007. Copyright 2009 by the American Academy of Pediatrics. All rights reserved. Print ISSN: 0031-4005. Online ISSN: 1098-4275.

Downloaded from pediatrics.aappublications.org at Indonesia:AAP Sponsored on December 5, 2011