Abstract
Although several studies have found no change or a decreased risk of childhood cancer in twins, few have controlled for potential confounders such as birth weight. We examined the association of birth plurality and childhood cancer in pooled data from five U.S. states (California, Minnesota, New York, Texas, and Washington) using linked birth-cancer registry data. The data, excluding children with Down syndrome or who died before 28 days of life, included 17,672 cases diagnosed from 1980 to 2004 at ages 28 days to 14 years and 57,966 controls with all cases and controls born from 1970 to 2004. Analyses were restricted to children weighing ≤4,000 g at birth. Odds ratios (OR) and 95% confidence intervals (95% CI) were estimated using unconditional logistic regression adjusting for sex, gestational age, birth weight, birth order, maternal age, maternal race, state of birth, and birth year. Children who were multiples had no difference in risk of cancer overall (OR, 0.93; 95% CI, 0.82-1.07), but a borderline reduced risk of Wilms' tumor (OR, 0.65; 95% CI, 0.39-1.09). For children diagnosed <2 y of age there was a reduced risk of Wilms' tumor (OR, 0.27; 95% CI, 0.09-0.86) and neuroblastoma (OR, 0.46; 95% CI, 0.25-0.84) and an increased risk of fibrosarcoma (OR, 5.81; 95% CI, 1.53-22.11). Higher-order multiple birth (triplets or higher) was not associated with childhood cancer. Our analysis suggests that mechanisms other than birth weight and gestational age may influence the lower risk of Wilms' tumor and neuroblastoma in multiple births. (Cancer Epidemiol Biomarkers Prev 2009;18(1):162–8)
Introduction
As many childhood cancers may initiate in utero (1), it is of interest to examine the role of pregnancy characteristics, including plurality, in the etiology of these diverse tumors. Plurality is also of timely importance because the rate of multiple births has risen dramatically in recent decades, from 19.3 per 1,000 births in 1980 to 33.8 per 1,000 births in 2005, largely due to increasing use of infertility treatment (2). Studies examining the association between twinship and childhood cancer risk have reported either null or decreased risk in twins compared with singletons (3-11). However, because these investigations have generally presented only unadjusted risk estimates, it is not clear whether these inverse associations reflect the influence of confounding variables or the unique biology of twin pregnancies. There are several maternal and birth factors that could influence both multiple birth and childhood cancer such as birth weight, maternal age, maternal race, and nutrition during pregnancy (12). For example, birth weight >4,000 g is positively associated with both acute lymphoblastic leukemia and Wilms' tumor (13, 14). Because infants of multiple pregnancies are generally smaller than singletons (15), reports of reduced risk of these tumors in twins may be confounded by birth weight.
Cancer risk among higher-order multiples (i.e. triplets or higher) has not been studied due to the rarity of both these pregnancies and childhood cancer. Higher-order multiple births are of particular recent interest because an increasing proportion of these children are born after the use of assisted reproductive technology (ART). It is estimated that 40% of infants in a triplet or higher birth were born following ART in 2004 in the United States, in contrast to only 17% of twins and 0.6% of singletons (16). ART procedures have also been hypothesized to increase the risk of cancer in offspring (17). Although several studies have failed to show an excess incidence of cancer following conception by ART (18-22), fertility treatment in general has been associated in single studies with hepatoblastoma (23), retinoblastoma (24), and imprinting disorders that predispose to several embryonal tumors (25). Examination of cancer among higher order multiples may inform ongoing investigation into the sequelae of ART.
In order to more precisely assess the role of birth characteristics in childhood cancer etiology, we pooled data from five states – California, Minnesota, New York (excluding New York City), Texas, and Washington – which had previously compiled case-control datasets by linking their cancer and birth registries (23, 26-29). The combined data set is among the largest to examine childhood cancer among twins and is the first, to our knowledge, to examine higher-order multiples.
Materials and Methods
Study Population
Approvals for the study were obtained from the institutional review boards at all participating institutions including each state's health department. Data from each state were deidentified prior to the pooled analysis and, as such, did not require consent of the subjects. Each state selected cases and controls in different years and with different eligibility criteria. The number of controls randomly selected from each state's birth registry varied from 1 control per case up to 10 controls per case. The matching criteria also varied by state. Minnesota, New York, Texas, and Washington used frequency-matching whereas California used individual matching. All states matched on birth year whereas two states matched on month of birth and sex as well. The methods of each state have been published elsewhere (23, 26-29) and are briefly detailed in Table 1. Individual observations from each state were combined into one dataset for analysis.
State . | Ages at diagnoses . | Years of diagnosis . | Years of birth . | n cases included . | n controls included . | Matching factors . |
---|---|---|---|---|---|---|
California | 28 d-4 y | 1988-1997 | 1983-1997 | 4,177 | 8,730 | Birth year, sex |
Minnesota | 28 d-14 y | 1988-2004 | 1976-2004 | 2,170 | 8,735 | Birth year |
New York | 28 d-14 y | 1985-2001 | 1970-2001 | 4,357 | 12,041 | Birth year |
Texas | 28 d-14 y | 1990-1998 | 1975-1998 | 4,647 | 4,732 | Birth year, sex |
Washington | 28 d-14 y | 1980-2004 | 1980-2004 | 2,321 | 23,728 | Birth year |
State . | Ages at diagnoses . | Years of diagnosis . | Years of birth . | n cases included . | n controls included . | Matching factors . |
---|---|---|---|---|---|---|
California | 28 d-4 y | 1988-1997 | 1983-1997 | 4,177 | 8,730 | Birth year, sex |
Minnesota | 28 d-14 y | 1988-2004 | 1976-2004 | 2,170 | 8,735 | Birth year |
New York | 28 d-14 y | 1985-2001 | 1970-2001 | 4,357 | 12,041 | Birth year |
Texas | 28 d-14 y | 1990-1998 | 1975-1998 | 4,647 | 4,732 | Birth year, sex |
Washington | 28 d-14 y | 1980-2004 | 1980-2004 | 2,321 | 23,728 | Birth year |
To facilitate pooling, additional criteria were applied to data from individual states. Because some states eliminated children diagnosed between birth and 28 d, these criteria were applied to all states. In addition, due to different sampling methods, some cancer cases were included in the control set in New York and Minnesota; these were excluded. In all analyses, children who had been recorded as having Down syndrome on their birth certificate were excluded (n = 100); however, due to reporting limitations (Down syndrome not recorded in Texas before 1984 and not in Washington before 1989), some children with Down syndrome identified at birth could still be included in the pooled dataset. In California, 73 controls who died before their matched case was diagnosed were replaced.
Specific case information included cancer diagnosis, age at diagnosis, and laterality for Wilms' tumor and retinoblastoma. Classification of cases was done using the International Classification of Childhood Cancer, Third Edition (ICCC-3; ref. 30). States that classified cases using the second edition (ICCC-2) codes were recoded to ICCC-3 codes for analysis.
Gestational age could be either calculated from last menstrual period or based on a clinical estimate. California provided the last menstrual period estimate only, whereas the other states provide both estimates of gestational age. For this analysis, if the last menstrual period estimate was available it was used; otherwise, the clinical estimate was used. Gestational ages <20 wk or >45 wk were considered implausible and were treated as missing, as were birth weights <350 g.
The primary analysis for this study was limited to children weighing ≤4,000 g at birth, because birth weights of multiples and singletons varied considerably. Adjustment for birth weight among children weighing >4,000 g would have included only seven twins and no children from higher-order multiple births. The pooled dataset included 17,672 cases and 57,966 controls, 15,068 and 50,286 of whom, respectively, weighed ≤4,000 g at birth. Among cases, 338 children were twins, and 14 were of triplets or higher-level multiples; among controls there were 1,252 twins and 41 triplets/higher-level multiples.
Statistical Analysis
Unconditional logistic regression was used to examine the association between birth plurality and childhood cancer. Individual matching in the California dataset was broken for the purpose of pooled analysis. Only cancers or broad cancer groups with at least four cases in each category of plurality were analyzed (at least two cases for subgroup analysis). Birth plurality was categorized in two ways: as a dichotomous variable (singleton versus multiple) and as a three-level variable (singleton, twin, and triplet or higher). Crude models were adjusted for pooling by controlling for sex, state of birth, and year of birth (by quartile). In the analysis of plurality as a dichotomous variable, birth weight was classified as <2,000 g, 2,000-2,499 g, 2,500-2,999 g, or ≥3,000 g, and gestational age was classified as <37 wk or ≥ 37 wk. In the analysis using the three-level plurality, variable birth weight was classified as <1,500 g or ≥1,500 g, and gestational age was classified as <28 wk or ≥28 wk because all but three higher order multiples had birth weights <2,500 g. In addition to adjustment for birth weight and gestational age, we also controlled for maternal age (<25, 25-29, 30-34, ≥35 y), maternal race (white, non-white), birth order (1st, 2nd, 3rd or higher), sex (male, female), state of birth, and birth year (by quartile). Size for gestational age was also examined as a covariate, with subjects classified as small (SGA, <10th percentile), appropriate (AGA, 10th-90th percentile), and large (LGA, >90th) based on the distribution among controls. Because no multiples were LGA, the AGA and LGA categories were collapsed. Models that included size for gestational age did not adjust for birth weight or gestational age.
Additional analyses were conducted with cases stratified a priori by age at diagnosis (0-4, 5-9, and 10-14 y) for all cancers combined, leukemia, lymphoma, and central nervous system tumors. Cases of embryonal tumors (neuroblastoma, retinoblastoma, Wilms' tumor, hepatoblastoma, intracranial embryonal central nervous system, and rhabdomyosarcoma), which occur mostly in the first 5 y of life, were divided a priori into those diagnosed at <2 and ≥ 2 y. Controls were included in these subgroup analyses when they could have been captured by the state's cancer registry had they been diagnosed with cancer based on the age range of the cases. We also analyzed the data restricting cases and controls to those born after 1989 because this was the median birth year and ART use became more frequent in later years. A final sensitivity analysis was done to assess the impact of different case-control ratios using only one control per case for all states. Homogeneity between the individual state estimates was assessed for the more common types of cancer using Wald χ2 tests. No indication of heterogeneity was detected. All analysis was done using SAS version 9.1 (SAS Institute).
Results
Descriptive characteristics of cases and controls by potential confounders are presented in Table 2. In the logistic regression analysis crude and adjusted estimates were fairly similar, with neither suggesting an association between childhood cancer overall and multiple birth [Crude odds ratio (OR), 0.94; 95% confidence interval (95% CI), 0.83-1.06; Adjusted OR, 0.93; 95% CI, 0.82-1.07; Table 3]. A significantly increased crude OR for hepatoblastoma among multiples (OR, 2.61; 95% CI, 1.56-4.35) was no longer present after controlling for potential confounders (OR, 0.92; 95% CI, 0.51-1.66).
Variable . | Category . | Cases (%) . | Controls (%) . | |||
---|---|---|---|---|---|---|
Mother's age at birth, y | <25 | 5,496 (36.5) | 19,144 (38.1) | |||
25-29 | 4,796 (31.8) | 15,914 (31.7) | ||||
30-34 | 3,260 (21.6) | 10,593 (21.1) | ||||
≥35 | 1,514 (10.1) | 4,619 (9.2) | ||||
Missing data | 2 | 16 | ||||
Mother's race | White | 13,031 (86.9) | 41,905 (84.2) | |||
Non-white | 1,962 (13.1) | 7,857 (15.8) | ||||
Missing data | 75 | 524 | ||||
Sex | Male | 8,030 (53.3) | 25,680 (51.1) | |||
Female | 7,037 (46.7) | 24,597 (48.9) | ||||
Missing data | 1 | 9 | ||||
Birth weight, g | <2,000 | 359 (2.4) | 1,045 (2.1) | |||
2,000-2,499 | 639 (4.2) | 2,108 (4.2) | ||||
2,500-2,999 | 2,396 (15.9) | 8,103 (16.1) | ||||
≥3,000 | 11,674 (77.5) | 39,030 (77.6) | ||||
Missing data | 0 | 0 | ||||
Gestational age, wk | <37 | 1,418 (9.8) | 4,258 (8.9) | |||
≥37 | 13,060 (90.2) | 43,774 (91.1) | ||||
Missing | 590 | 2,254 | ||||
Plurality | Singleton | 14,714 (97.7) | 48,988 (97.4) | |||
Twin | 338 (2.2) | 1,252 (2.5) | ||||
Triplet or higher | 14 (0.1) | 41 (0.1) | ||||
Missing data | 2 | 5 | ||||
Birth order | First | 6,320 (42.6) | 20,645 (42.1) | |||
Second | 4,803 (32.4) | 15,800 (32.3) | ||||
Third or higher | 3,715 (25.0) | 12,552 (25.6) | ||||
Missing | 230 | 1,289 | ||||
Size for gestational age | ||||||
Small | 1,604 (11.1) | 5,365 (11.2) | ||||
Average | 12,616 (87.1) | 41,989 (87.4) | ||||
Large | 258 (1.8) | 676 (1.4) | ||||
Missing data | 590 | 2,256 | ||||
Birth year | 1970-1985 | 3,953 (26.2) | 14,841 (29.5) | |||
1986-1989 | 3,984 (26.4) | 11,137 (22.1) | ||||
1990-1993 | 4,085 (27.1) | 11,763 (23.4) | ||||
1994-2004 | 3,046 (20.2) | 12,545 (24.9) |
Variable . | Category . | Cases (%) . | Controls (%) . | |||
---|---|---|---|---|---|---|
Mother's age at birth, y | <25 | 5,496 (36.5) | 19,144 (38.1) | |||
25-29 | 4,796 (31.8) | 15,914 (31.7) | ||||
30-34 | 3,260 (21.6) | 10,593 (21.1) | ||||
≥35 | 1,514 (10.1) | 4,619 (9.2) | ||||
Missing data | 2 | 16 | ||||
Mother's race | White | 13,031 (86.9) | 41,905 (84.2) | |||
Non-white | 1,962 (13.1) | 7,857 (15.8) | ||||
Missing data | 75 | 524 | ||||
Sex | Male | 8,030 (53.3) | 25,680 (51.1) | |||
Female | 7,037 (46.7) | 24,597 (48.9) | ||||
Missing data | 1 | 9 | ||||
Birth weight, g | <2,000 | 359 (2.4) | 1,045 (2.1) | |||
2,000-2,499 | 639 (4.2) | 2,108 (4.2) | ||||
2,500-2,999 | 2,396 (15.9) | 8,103 (16.1) | ||||
≥3,000 | 11,674 (77.5) | 39,030 (77.6) | ||||
Missing data | 0 | 0 | ||||
Gestational age, wk | <37 | 1,418 (9.8) | 4,258 (8.9) | |||
≥37 | 13,060 (90.2) | 43,774 (91.1) | ||||
Missing | 590 | 2,254 | ||||
Plurality | Singleton | 14,714 (97.7) | 48,988 (97.4) | |||
Twin | 338 (2.2) | 1,252 (2.5) | ||||
Triplet or higher | 14 (0.1) | 41 (0.1) | ||||
Missing data | 2 | 5 | ||||
Birth order | First | 6,320 (42.6) | 20,645 (42.1) | |||
Second | 4,803 (32.4) | 15,800 (32.3) | ||||
Third or higher | 3,715 (25.0) | 12,552 (25.6) | ||||
Missing | 230 | 1,289 | ||||
Size for gestational age | ||||||
Small | 1,604 (11.1) | 5,365 (11.2) | ||||
Average | 12,616 (87.1) | 41,989 (87.4) | ||||
Large | 258 (1.8) | 676 (1.4) | ||||
Missing data | 590 | 2,256 | ||||
Birth year | 1970-1985 | 3,953 (26.2) | 14,841 (29.5) | |||
1986-1989 | 3,984 (26.4) | 11,137 (22.1) | ||||
1990-1993 | 4,085 (27.1) | 11,763 (23.4) | ||||
1994-2004 | 3,046 (20.2) | 12,545 (24.9) |
Cancer type . | Any multiple versus. singletons . | . | . | . | |||
---|---|---|---|---|---|---|---|
. | Singleton n (%) . | Multiple n (%) . | Crude OR* (95% CI) . | OR† (95% CI) . | |||
Controls | 48,988 (97.4) | 1,293 (2.6) | Ref | Ref | |||
All cancers | 14,714 (97.7) | 352 (2.3) | 0.94 (0.83-1.06) | 0.93 (0.82-1.07) | |||
Leukemia | 4,915 (97.7) | 118 (2.3) | 0.96 (0.79-1.16) | 1.04 (0.85-1.29) | |||
ALL | 3,962 (97.7) | 93 (2.3) | 0.94 (0.75-1.16) | 1.04 (0.82-1.32) | |||
AML | 705 (97.5) | 18 (2.5) | 0.99 (0.62-1.59) | 0.96 (0.58-1.61) | |||
Lymphoma | 1,271 (97.8) | 29 (2.2) | 0.90 (0.62-1.30) | 0.91 (0.61-1.36) | |||
HD | 402 (97.3) | 11 (2.7) | 1.15 (0.64-2.06) | 1.30 (0.69-2.48) | |||
NHL | 503 (98.2) | 9 (1.8) | 0.69 (0.35-1.33) | 0.61 (0.29-1.26) | |||
Central nervous system tumors | 3,171 (97.8) | 71 (2.2) | 0.88 (0.69-1.12) | 0.91 (0.70-1.19) | |||
Ependymoma | 326 (98.2) | 6 (1.8) | 0.71 (0.32-1.60) | 0.96 (0.41-2.23) | |||
Astrocytoma | 1,405 (98.1) | 27 (1.9) | 0.77 (0.53-1.13) | 0.86 (0.57-1.29) | |||
Intracranial embryonal | 741 (97.5) | 19 (2.5) | 1.00 (0.63-1.59) | 1.00 (0.62-1.64) | |||
Other gliomas | 424 (97.9) | 9 (2.1) | 0.85 (0.44-1.64) | 0.63 (0.30-1.31) | |||
Embryonal tumors‡ | 4,273 (97.5) | 111 (2.5) | 1.00 (0.82-1.22) | 0.90 (0.73-1.12) | |||
Neuroblastoma | 1,252 (97.8) | 28 (2.2) | 0.85 (0.58-1.24) | 0.75 (0.50-1.12) | |||
Retinoblastoma | 588 (96.9) | 19 (3.1) | 1.24 (0.78-1.98) | 1.17 (0.70-1.94) | |||
Wilms' tumor | 980 (98.4) | 16 (1.6) | 0.63 (0.38-1.04) | 0.65 (0.39-1.09) | |||
Hepatoblastoma | 227 (93.4) | 16 (6.6) | 2.60 (1.56-4.35) | 0.92 (0.51-1.66) | |||
Bone tumors | 476 (97.9) | 10 (2.1) | 0.83 (0.44-1.57) | 0.92 (0.48-1.79) | |||
Osteosarcoma | 233 (97.5) | 6 (2.5) | 0.99 (0.44-2.25) | 1.29 (0.55-3.05) | |||
Soft tissue sarcoma | 874 (97.0) | 27 (3.0) | 1.23 (0.84-1.81) | 1.42 (0.94-2.15) | |||
Rhabdomyosarcoma | 485 (97.4) | 13 (2.6) | 1.03 (0.59-1.80) | 1.33 (0.74-2.40) | |||
Fibrosarcoma | 104 (95.4) | 5 (4.6) | 1.79 (0.73-4.41) | 1.64 (0.61-4.42) | |||
Germ cell tumors | 474 (98.3) | 8 (1.7) | 0.65 (0.32-1.31) | 0.73 (0.35-1.52) | |||
Gonadal germ cell tumors | 231 (98.3) | 4 (1.7) | 0.67 (0.25-1.79) | 0.70 (0.25-1.97) | |||
Carcinomas | 345 (97.7) | 8 (2.3) | 0.91 (0.45-1.85) | 0.69 (0.30-1.61) | |||
Thyroid carcinoma | 138 (96.5) | 5 (3.5) | 1.44 (0.59-3.53) | 0.95 (0.28-3.22) |
Cancer type . | Any multiple versus. singletons . | . | . | . | |||
---|---|---|---|---|---|---|---|
. | Singleton n (%) . | Multiple n (%) . | Crude OR* (95% CI) . | OR† (95% CI) . | |||
Controls | 48,988 (97.4) | 1,293 (2.6) | Ref | Ref | |||
All cancers | 14,714 (97.7) | 352 (2.3) | 0.94 (0.83-1.06) | 0.93 (0.82-1.07) | |||
Leukemia | 4,915 (97.7) | 118 (2.3) | 0.96 (0.79-1.16) | 1.04 (0.85-1.29) | |||
ALL | 3,962 (97.7) | 93 (2.3) | 0.94 (0.75-1.16) | 1.04 (0.82-1.32) | |||
AML | 705 (97.5) | 18 (2.5) | 0.99 (0.62-1.59) | 0.96 (0.58-1.61) | |||
Lymphoma | 1,271 (97.8) | 29 (2.2) | 0.90 (0.62-1.30) | 0.91 (0.61-1.36) | |||
HD | 402 (97.3) | 11 (2.7) | 1.15 (0.64-2.06) | 1.30 (0.69-2.48) | |||
NHL | 503 (98.2) | 9 (1.8) | 0.69 (0.35-1.33) | 0.61 (0.29-1.26) | |||
Central nervous system tumors | 3,171 (97.8) | 71 (2.2) | 0.88 (0.69-1.12) | 0.91 (0.70-1.19) | |||
Ependymoma | 326 (98.2) | 6 (1.8) | 0.71 (0.32-1.60) | 0.96 (0.41-2.23) | |||
Astrocytoma | 1,405 (98.1) | 27 (1.9) | 0.77 (0.53-1.13) | 0.86 (0.57-1.29) | |||
Intracranial embryonal | 741 (97.5) | 19 (2.5) | 1.00 (0.63-1.59) | 1.00 (0.62-1.64) | |||
Other gliomas | 424 (97.9) | 9 (2.1) | 0.85 (0.44-1.64) | 0.63 (0.30-1.31) | |||
Embryonal tumors‡ | 4,273 (97.5) | 111 (2.5) | 1.00 (0.82-1.22) | 0.90 (0.73-1.12) | |||
Neuroblastoma | 1,252 (97.8) | 28 (2.2) | 0.85 (0.58-1.24) | 0.75 (0.50-1.12) | |||
Retinoblastoma | 588 (96.9) | 19 (3.1) | 1.24 (0.78-1.98) | 1.17 (0.70-1.94) | |||
Wilms' tumor | 980 (98.4) | 16 (1.6) | 0.63 (0.38-1.04) | 0.65 (0.39-1.09) | |||
Hepatoblastoma | 227 (93.4) | 16 (6.6) | 2.60 (1.56-4.35) | 0.92 (0.51-1.66) | |||
Bone tumors | 476 (97.9) | 10 (2.1) | 0.83 (0.44-1.57) | 0.92 (0.48-1.79) | |||
Osteosarcoma | 233 (97.5) | 6 (2.5) | 0.99 (0.44-2.25) | 1.29 (0.55-3.05) | |||
Soft tissue sarcoma | 874 (97.0) | 27 (3.0) | 1.23 (0.84-1.81) | 1.42 (0.94-2.15) | |||
Rhabdomyosarcoma | 485 (97.4) | 13 (2.6) | 1.03 (0.59-1.80) | 1.33 (0.74-2.40) | |||
Fibrosarcoma | 104 (95.4) | 5 (4.6) | 1.79 (0.73-4.41) | 1.64 (0.61-4.42) | |||
Germ cell tumors | 474 (98.3) | 8 (1.7) | 0.65 (0.32-1.31) | 0.73 (0.35-1.52) | |||
Gonadal germ cell tumors | 231 (98.3) | 4 (1.7) | 0.67 (0.25-1.79) | 0.70 (0.25-1.97) | |||
Carcinomas | 345 (97.7) | 8 (2.3) | 0.91 (0.45-1.85) | 0.69 (0.30-1.61) | |||
Thyroid carcinoma | 138 (96.5) | 5 (3.5) | 1.44 (0.59-3.53) | 0.95 (0.28-3.22) |
Abbreviations: ALL, acute lymphoblastic leukemia; AML, acute myeloid leukemia; HD, Hodgkin's disease; NHL, non-Hodgkin's lymphoma.
Adjusted only for matching variables; state, sex, and birth year.
Adjusted for maternal race, maternal age, sex, state, gestational age, birth year, birth order, and birth weight.
Includes neuroblastoma, retinoblastoma, Wilms' tumor, hepatoblastoma, intracranial embryonal central nervous system, and rhabdomyosarcoma.
In models adjusting for potential confounders, there was a marginally reduced risk of Wilms' tumor in children born as part of a multiple birth for all age groups (OR, 0.65; 95% CI, 0.39-1.09). No associations of other cancers with multiple births were identified in the overall analysis. In age-specific analysis, there was a reduced risk of embryonal tumors combined (OR, 0.64; 95% CI, 0.46-0.90), and specifically for Wilms' tumor (OR, 0.27; 95% CI, 0.09-0.86) and neuroblastoma (OR, 0.46; 95% CI, 0.25-0.84) for cases diagnosed <2 years of age (Table 4). There was also an increased risk of soft tissue sarcoma overall (OR, 2.30; 95% CI, 1.12-4.69) and specifically fibrosarcoma (OR, 5.81; 95% CI, 1.53-22.11) in the younger age group.
Cancer type . | Any multiple versus singletons . | . | . | . | . | . | ||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
. | Age 0-4 y . | . | Age 5-9 y . | . | Age 10-14 y . | . | ||||||
. | Multiple n (%) . | OR* (95% CI) . | Multiple n (%) . | OR* (95% CI) . | Multiple n (%) . | OR* (95% CI) . | ||||||
All cancers | 227 (2.4) | 0.92 (0.78-1.08) | 75 (2.6) | 1.07 (0.82-1.39) | 50 (1.9) | 0.80 (0.57-1.12) | ||||||
Leukemia | 78 (2.2) | 0.96 (90.75-1.24) | 29 (3.0) | 1.37 (0.91-2.06) | 11 (1.9) | 0.83 (0.41-1.68) | ||||||
ALL | 66 (2.3) | 1.01 (0.77-1.33) | 19 (2.4) | 1.12 (0.69-1.83) | 8 (2.0) | 0.92 (0.39-2.15) | ||||||
AML | 10 (2.1) | 0.77 (0.38-1.55) | 6 (5.0) | 1.94 (0.78-4.84) | 2 (1.6) | 0.66 (0.15-2.83) | ||||||
Lymphoma | 12 (2.9) | 0.96 (0.52-1.77) | 6(1.8) | 0.69 (0.30-1.59) | 11 (2.0) | 1.02 (0.53-1.98) | ||||||
HD | 2 (4.7) | 1.30 (0.27-6.23) | 2 (2.4) | 0.70 (0.16-3.02) | 7 (2.4) | 1.70 (0.76-3.80) | ||||||
NHL | 5 (2.7) | 0.96 (0.37-2.48) | 1 (0.6) | † | 3 (1.8) | 0.53 (0.13-2.24) | ||||||
Central nervous system tumors | 38 (2.2) | 0.90 (0.63-1.29) | 23 (2.5) | 1.02 (0.64-1.61) | 10 (1.6) | 0.76 (0.38-1.54) | ||||||
Ependymoma | 4 (1.7) | 0.90 (0.32-2.53) | 2 (3.6) | 1.78 (0.40-8.00) | 0 (0.0) | † | ||||||
Astrocytoma | 17 (2.5) | 1.14 (0.68-1.91) | 6 (1.4) | 0.62 (0.27-1.44) | 4 (1.3) | 0.51 (0.16-1.65) | ||||||
Intracranial embryonal | 6 (1.3) | 0.52 (0.23-1.20) | 9 (4.7) | 1.57 (0.75-3.28) | 4 (3.7) | 2.28 (0.77-6.76) | ||||||
Other gliomas | 5 (2.4) | 0.57 (0.20-1.63) | 4 (2.6) | 1.00 (0.35-2.91) | 0 (0.0) | † | ||||||
Age <2 y | Age ≥2 y | |||||||||||
Embryonal tumors‡ | 45 (2.3) | 0.64 (0.46-0.90) | 66 (2.7) | 1.16 (0.89-1.52) | ||||||||
Neuroblastoma | 13 (1.8) | 0.46 (0.25-0.84) | 15 (2.8) | 1.33 (0.76-2.30) | ||||||||
Retinoblastoma | 13 (3.4) | 1.17 (0.63-2.19) | 6 (2.7) | 1.10 (0.46-2.62) | ||||||||
Wilms tumor | 3 (0.8) | 0.27 (0.09-0.86) | 13 (2.1) | 0.93 (0.52-1.66) | ||||||||
Hepatoblastoma | 11 (6.4) | 0.80 (0.39-1.64) | 5 (7.1) | 1.25 (0.45-3.50) | ||||||||
Soft tissue sarcoma | 10 (4.9) | 2.30 (1.12-4.69) | 17 (2.4) | 1.16 (0.69-1.93) | ||||||||
Rhabdomyosarcoma | 3 (2.6) | 1.45 (0.43-4.93) | 10 (2.6) | 1.29 (0.66-2.51) | ||||||||
Fibrosarcoma | 4 (12.9) | 5.81 (1.53-22.11) | 1 (1.3) | † |
Cancer type . | Any multiple versus singletons . | . | . | . | . | . | ||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
. | Age 0-4 y . | . | Age 5-9 y . | . | Age 10-14 y . | . | ||||||
. | Multiple n (%) . | OR* (95% CI) . | Multiple n (%) . | OR* (95% CI) . | Multiple n (%) . | OR* (95% CI) . | ||||||
All cancers | 227 (2.4) | 0.92 (0.78-1.08) | 75 (2.6) | 1.07 (0.82-1.39) | 50 (1.9) | 0.80 (0.57-1.12) | ||||||
Leukemia | 78 (2.2) | 0.96 (90.75-1.24) | 29 (3.0) | 1.37 (0.91-2.06) | 11 (1.9) | 0.83 (0.41-1.68) | ||||||
ALL | 66 (2.3) | 1.01 (0.77-1.33) | 19 (2.4) | 1.12 (0.69-1.83) | 8 (2.0) | 0.92 (0.39-2.15) | ||||||
AML | 10 (2.1) | 0.77 (0.38-1.55) | 6 (5.0) | 1.94 (0.78-4.84) | 2 (1.6) | 0.66 (0.15-2.83) | ||||||
Lymphoma | 12 (2.9) | 0.96 (0.52-1.77) | 6(1.8) | 0.69 (0.30-1.59) | 11 (2.0) | 1.02 (0.53-1.98) | ||||||
HD | 2 (4.7) | 1.30 (0.27-6.23) | 2 (2.4) | 0.70 (0.16-3.02) | 7 (2.4) | 1.70 (0.76-3.80) | ||||||
NHL | 5 (2.7) | 0.96 (0.37-2.48) | 1 (0.6) | † | 3 (1.8) | 0.53 (0.13-2.24) | ||||||
Central nervous system tumors | 38 (2.2) | 0.90 (0.63-1.29) | 23 (2.5) | 1.02 (0.64-1.61) | 10 (1.6) | 0.76 (0.38-1.54) | ||||||
Ependymoma | 4 (1.7) | 0.90 (0.32-2.53) | 2 (3.6) | 1.78 (0.40-8.00) | 0 (0.0) | † | ||||||
Astrocytoma | 17 (2.5) | 1.14 (0.68-1.91) | 6 (1.4) | 0.62 (0.27-1.44) | 4 (1.3) | 0.51 (0.16-1.65) | ||||||
Intracranial embryonal | 6 (1.3) | 0.52 (0.23-1.20) | 9 (4.7) | 1.57 (0.75-3.28) | 4 (3.7) | 2.28 (0.77-6.76) | ||||||
Other gliomas | 5 (2.4) | 0.57 (0.20-1.63) | 4 (2.6) | 1.00 (0.35-2.91) | 0 (0.0) | † | ||||||
Age <2 y | Age ≥2 y | |||||||||||
Embryonal tumors‡ | 45 (2.3) | 0.64 (0.46-0.90) | 66 (2.7) | 1.16 (0.89-1.52) | ||||||||
Neuroblastoma | 13 (1.8) | 0.46 (0.25-0.84) | 15 (2.8) | 1.33 (0.76-2.30) | ||||||||
Retinoblastoma | 13 (3.4) | 1.17 (0.63-2.19) | 6 (2.7) | 1.10 (0.46-2.62) | ||||||||
Wilms tumor | 3 (0.8) | 0.27 (0.09-0.86) | 13 (2.1) | 0.93 (0.52-1.66) | ||||||||
Hepatoblastoma | 11 (6.4) | 0.80 (0.39-1.64) | 5 (7.1) | 1.25 (0.45-3.50) | ||||||||
Soft tissue sarcoma | 10 (4.9) | 2.30 (1.12-4.69) | 17 (2.4) | 1.16 (0.69-1.93) | ||||||||
Rhabdomyosarcoma | 3 (2.6) | 1.45 (0.43-4.93) | 10 (2.6) | 1.29 (0.66-2.51) | ||||||||
Fibrosarcoma | 4 (12.9) | 5.81 (1.53-22.11) | 1 (1.3) | † |
Adjusted for maternal race, maternal age, sex, state, gestational age, birth year, birth order, and birth weight; ORs not reported if <2 exposed cases.
One or fewer cases of twins or higher in the case set.
Includes neuroblastoma, retinoblastoma, Wilms tumor, hepatoblastoma, intracranial embryonal CNS, and rhabdomyosarcoma.
Results of analyses controlling for size for gestational age (data not shown) were similar to those adjusting separately for birth weight and gestational age with the exception of hepatoblastoma in all ages combined and neuroblastoma in children diagnosed before the age of 2 years. The association with hepatoblastoma was significant and close to the crude model estimate (OR, 2.30; 95% CI, 1.33-3.99) whereas the association with neuroblastoma was attenuated and became only marginally significant (OR, 0.59; 95% CI, 0.33-1.06).
The analysis limited to children born after 1989, the median birth year past which the prevalence of ART use is expected to be higher, gave results similar to the overall analysis (data not shown). However, there was an increased risk of soft tissue sarcoma overall (OR, 2.00; 95% CI, 1.13-3.55) and specifically fibrosarcoma (OR, 4.22; 95% CI, 1.22-14.60). This analysis is similar to the analysis in the children <2 years of age as it includes the same fibrosarcoma cases.
Few children were part of a higher-order multiple birth, and there were no significant associations with childhood cancer overall or with broad cancer groups with at least four cases who were members of triplet or higher births (Table 5). Ten of the 14 cases who were part of a triplet or higher birth were born after 1989, as were 25 of 41 controls. Limiting the data to birth years after 1989, the ORs increased but remained nonsignificant for cancer overall (OR, 1.34; 95% CI, 0.61-2.95) and for embryonal tumors (OR, 1.68; 95% CI, 0.63-4.47), most of which were hepatoblastoma cases. Central nervous system tumors could not be analyzed among the latter period because only one case who was part of a higher order birth was observed.
Cancer type . | Twins vs. singletons . | . | Triplets or higher vs. singletons . | . | ||
---|---|---|---|---|---|---|
. | n (%) . | OR* (95% CI) . | n (%) . | OR* (95% CI) . | ||
Controls | 1,252 (2.5) | Ref | 41 (0.08) | Ref | ||
All cancers | 338 (2.2) | 0.93 (0.81-1.06) | 14 (0.09) | 1.12 (0.59-2.13) | ||
Central nervous system tumors | 67 (2.1) | 0.88 (0.68-1.14) | 4 (0.12) | 1.62 (0.56-4.67) | ||
Embryonal tumors† | 105 (2.4) | 0.95 (0.77-1.17) | 6 (0.14)‡ | 1.19 (0.48-2.97) |
Cancer type . | Twins vs. singletons . | . | Triplets or higher vs. singletons . | . | ||
---|---|---|---|---|---|---|
. | n (%) . | OR* (95% CI) . | n (%) . | OR* (95% CI) . | ||
Controls | 1,252 (2.5) | Ref | 41 (0.08) | Ref | ||
All cancers | 338 (2.2) | 0.93 (0.81-1.06) | 14 (0.09) | 1.12 (0.59-2.13) | ||
Central nervous system tumors | 67 (2.1) | 0.88 (0.68-1.14) | 4 (0.12) | 1.62 (0.56-4.67) | ||
Embryonal tumors† | 105 (2.4) | 0.95 (0.77-1.17) | 6 (0.14)‡ | 1.19 (0.48-2.97) |
Adjusted for maternal race, maternal age, sex, state, gestational age, birth year, birth order, and birth weight.
Includes neuroblastoma, retinoblastoma, Wilms' tumor, hepatoblastoma, intracranial embryonal central nervous system, and rhabdomyosarcoma.
Four cases were hepatoblastoma, one was retinoblastoma, and one was Wilms' tumor.
ORs did not differ from those in the main analysis when the number of controls was limited to only one per case (data not shown).
Discussion
We found a suggestion of a decrease in the risk of Wilms' tumor for children born as part of a multiple birth for all age groups combined and a significantly decreased risk of Wilms' tumor and neuroblastoma in children <2 years. That these associations remained after controlling for a number of potential confounders suggests that mechanisms other than birth weight and gestational age may influence the risk of Wilms' tumor and neuroblastoma in multiple births. We also observed an increased risk of fibrosarcoma in children diagnosed before the age of 2 years. Higher-order multiple births were not associated with childhood cancer, although even in a dataset of this large size there were few children who were part of such births.
Several previous studies have examined the association between multiple births and childhood cancer, but have had varying ability to control for potential confounders and examine specific cancer diagnoses. For childhood cancer overall, unadjusted estimates from previous studies have indicated no difference or a small reduced risk in twins compared with singletons (3-11). Two previous studies of twinship and childhood cancer have attempted to control for birth weight. One study assessed whether cancer occurred more often in the heavier twin, which was the case for leukemia but not all other cancer types combined (5, 9). Another found that there was a reduced incidence of childhood cancer overall in twins weighing <3,000 g at birth and an excess in twins weighing ≥3,000 g (7). Among the few studies that have examined specific cancer diagnoses besides leukemia, one reported fewer cases of renal cancer than expected (4). A second study found a decreased risk for cancer overall, hematopoietic cancer (leukemia and lymphoma), and Wilms' tumor (6). A third study, by contrast, reported an increased risk of renal cancer in twins (11). Our results for Wilms' tumor were generally concordant with previous studies that examined renal tumors or Wilms' tumors specifically. Few studies have examined less common childhood cancer diagnoses, so there is little basis for comparison to prior research.
It has been proposed that the inverse association between certain childhood cancers and multiple birth may be due to intrauterine growth restriction rather than birth weight per se (6). We attempted to assess if intrauterine growth restriction, using size for gestational age, explained the associations seen in our study. Analyses that included size for gestational age as a measure for intrauterine growth restriction were similar to models adjusting for birth weight and gestational age as separate variables.
Other mechanisms might also be posited, however speculative, given that neither birth weight nor intrauterine growth restriction explained our observed associations. For example, some studies have found that high folic acid intake may increase the rate of twinning (31-33). Likewise, folic acid supplementation has been associated with a decreased risk of neuroblastoma (34-37). Other dietary, environmental, or intrinsic exposures that differ between multiple and singleton pregnancies should be considered as well.
Although our numbers were very limited, most of the children who were part of a triplet or higher birth were born after 1989, the median birth year in the dataset. Although the use of ART was still infrequent, these later birth years would be expected to have an increasing number of children born following conception by ART. All of the children with embryonal tumors who were part of a triplet or higher birth were born after 1989 whereas only one of the four multiples with central nervous system tumors was diagnosed in this later time period. Although this could represent a difference in age distribution in particular cancer diagnoses, it could also represent an increase in embryonal tumors among higher-order multiple births following the advent of ART procedures. However, as most of the embryonal cancers among higher-order multiples were hepatoblastoma, any apparent relationship might instead reflect an incompletely controlled association with low birth weight rather than multiple birth per se because very low birth weight is a well-established risk factor for hepatoblastoma (16, 23, 38-41).
The major strength of our study was the large number of childhood cancer cases. We had 80% power to detect an OR of 1.17 (or its inverse, 0.85) comparing multiple with singleton births for all childhood cancers combined; minimum detectable ORs ranged from 1.3 to 5.3 (0.8-0.1) for individual cancer diagnosis. For most diagnoses the minimum detectable OR was <2.5 (0.4). Few other studies have examined the relationship between multiple birth and childhood cancer after controlling for potentially confounding birth characteristics. In addition, with a large sample size we were able to examine less common childhood cancers and age-specific subgroups. Finally, the study was population-based; most cases of cancer in each state in the time periods selected were included along with a random sample of births.
Several limitations of this study should also be noted. No information was available on sex concordance, zygosity, or birth sequence. Thus, we could not separate the effects of shared fetal environment from that of shared genetic predisposition. We also may have sampled more than one member of a multiple birth, because no identifying information was available in this limited dataset to screen out siblings. We could identify potential multiple sibships by matching birth year, state of birth, parental age, parental race/ethnicity, and parental education. No higher-order births were found to be possible sibships, but there were seven possible twinships. Of these pairs, three were both controls, three were both cases (both Wilms' tumor, both myelodysplastic syndrome, and one rhabdomyosarcoma/ Wilms' tumor), and one set was a case (neuroblastoma) and a control. However, although these subjects each had common values for matching variables, there is no guarantee that they represented actual twinships.
Other limitations include multiple comparisons, potential measurement error based on birth certificate data, and the lack of data on fetal loss and infant death. With many cancer subtypes examined, significant results in the subgroup analyses in particular could be due to chance alone rather than reflecting a true association. Data items collected on birth certificates have varying levels of accuracy. However, most of the variables used in this analysis, such as maternal age, birth plurality, and birth weight, have been shown to be accurate in validation studies (42). Finally, information on fetal losses or mortality past 28 days of life was not collected, which could have affected our results in a number of ways. Multiple pregnancy is associated with an increase risk of spontaneous abortion and fetal reduction such that children who might have been twins initially were born as singletons (43). Because Wilms' tumor and neuroblastoma are known to develop prenatally (44, 45), it is possible that children who had or would have developed these cancers were more likely to die in utero, giving the appearance of a reduction in risk of cancer. However, leukemia, germ cell tumors, and other embryonal tumors also are thought to have prenatal origins, but did not show inverse associations with multiple births (1, 46-48). Twins also have five times the risk of death in the first year of life as do singletons (49), making competing risk more of a concern among multiple births. It is also possible that multiples who died in the first year would have been more likely to develop cancer, had they lived, than twins who survived the first year of life. However, if this were the case we might expect an inverse relationship in all cancers, rather than in Wilms' tumor and neuroblastoma alone.
In conclusion, ours is the first well-controlled study of childhood cancer in twins and higher-order multiple births. Potential mechanisms beyond birth weight and growth such as dietary, environmental, and genetic exposures should be investigated. These exposures could potentially explain the reduction in risk for Wilms' tumor and neuroblastoma, and possible increase in risk for fibrosarcoma, among children born as part of a multiple birth.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Grant support: Support for analysis of the pooled dataset was provided by the Children's Cancer Research Fund, Minneapolis, MN and National Institutes of Health Pediatric Cancer Epidemiology Training Grant T32 CA099936, as was support for the assembly of the Minnesota and Washington datasets. The Washington State Cancer Registry and the Cancer Surveillance System of Western Washington, which provided data, are supported by contract N01-CN-05230 from the National Cancer Institute and the Fred Hutchinson Cancer Research Center. In California and Texas, National Cancer Institute grants R01CA717450 and R01CA92670 supported assembly of their respective datasets. In New York, partial support for assembly of the dataset was received from the Centers for Disease Control and Prevention's National Program of Cancer Registries by cooperative agreement U58DP000783-01 awarded to the New York State Department of Health; contents of this manuscript are solely the responsibility of the authors and do not necessarily represent the official views of the Centers for Disease Control and Prevention.
Acknowledgments
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We acknowledge the programming assistance of Mr. Eric Elkin, Ms. Susan Hurley, Mr. John Soler, and Mr. Bill O'Brien, and thank the Washington State Department of Health, the Minnesota Cancer Surveillance System, and other collaborating institutions for allowing data access.