SEARCH

SEARCH BY CITATION

Keywords:

  • Wilms tumor;
  • birth characteristics;
  • childhood cancer;
  • case-cohort

Abstract

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Result
  5. Discussion
  6. References

Wilms tumor (WT) is a childhood kidney cancer with the highest rate of occurrence before the age of 2. Since it is rare, previous research has been limited and few risk factors have been established. We used a case-cohort design to examine the influence of birth characteristics on occurrence of WT in Minnesota. A total of 2,188 cases of cancer diagnosed in children aged 28 days to 14 years from 1988 to 2004 were identified using the Minnesota Cancer Surveillance System (MCSS). For each case, 4 children were randomly selected from Minnesota birth records during 1976–2004, frequency matched on birth year. Thus, a total of 8,752 children comprised the subcohort for the study, who in this analysis, served as comparison to the 138 cases of WT. Study variables included parental demographics, maternal pregnancy history and conditions and health and conditions of the child at birth. Associations with WT were assessed using hazard ratios (HR) and 95% confidence intervals (CI) calculated from stratified Cox regression models. We found an increased risk of WT for children who were large for gestational age compared to those average for gestational age and for children with congenital abnormalities. There was also an increased risk for children with a birth weight > 4,000 g compared to those with a birth weight between 2,500 and 4,000 g. All other factors examined showed no association with WT. This study contributes to the mounting evidence that children with large size at birth have an increased risk of WT. © 2007 Wiley-Liss, Inc.

Wilms tumor (WT)is an embryonal tumor that comprises 95% of renal cancer in children.1 In the United States, the age-adjusted annual incidence rate is 7.6 per million children, with the highest rate of tumors, 21 per million children, occurring within the first 2 years of life.1 The young age at diagnosis and poorly differentiated cells typical of WT suggest that it may arise during gestation.

Few risk factors for WT have been confirmed. Congenital genitourinary anomalies and aniridia are known to increase the risk of WT many fold, as are the “overgrowth” conditions such as Beckwith-Wiedemann syndrome and hemihypertrophy.2, 3 Environmental risk factors have been more difficult to research due to the rarity of WT. Several studies have found an increased risk of WT based on paternal occupational exposure to various chemicals including pesticides,4, 5 lead or hydrocarbons6–8 and boron.9 Other exogenous risk factors, such as maternal antibiotic use,10 household pest extermination,10 maternal use of hair coloring products,11 tea consumption,11 vaginal infections11 and penthrane exposure,12 have been associated with WT only in single studies.

Birth characteristics as risk factors for WT have been examined with some frequency. High birth weight has been most often studied, with several findings of an increased risk for WT for children with a birth weight above 4,000 g,13–16 although other studies have shown no association.10–12, 17, 18 Maternal hypertension,10–12, 15 advanced maternal age,10–12, 15, 17, 19, 20 Apgar score17 and pregnancy complications14 have demonstrated little or no evidence of association with WT.

We conducted a case-cohort study by linking the cancer and birth registries in Minnesota in order to examine the role of birth factors in the etiology of childhood cancers. Data were collected at birth, before cancer occurrence, so that results are robust to selection and recall biases. Although other studies with similarly strong designs have been performed in other countries, this is the first registry-based study of WT reported in a United States population.

Material and methods

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Result
  5. Discussion
  6. References

Cases of cancer diagnosed in children aged 28 days to 14 years from 1988 to 2004, inclusive, were identified using the Minnesota Cancer Surveillance System (MCSS) maintained by the Minnesota Department of Health. When there were multiple cancer occurrences, both were counted if individual diagnoses were less than 12 months apart with the assumption that treatment-related secondary cancers are unlikely within this timespan.21 There were 2 such cases out of all cancers. One child was diagnosed with neuroblastoma and WT on the same day and the second child was diagnosed with a central nervous system tumor 6 months after diagnosis with WT. Nearly all cases of cancer in the state of Minnesota are ascertained by MCSS; an external audit performed in 2002 estimated that 99.9% of all cancer cases in the state were included in the registry.22 Probabilistic record linkage was used to match cases to their birth certificate.23 Out of all the cancer cases identified by MCSS, 82.4% were successfully linked to their birth record. The Institutional Review Board (IRB) of the University of Minnesota, the IRB of the Minnesota Department of Health along with the MCSS have approved this study.

A comparison group was selected from birth certificates of all Minnesota children born from 1976 to 2004. For those born after 1980 only those who survived at least 28 days from birth were included, prior to 1980 neonatal deaths were not record and thus could not be excluded. For each case that was successfully linked to his or her birth record, 4 randomly selected birth certificates were chosen from the same birth year. The selected subcohort of births represents the exposure distribution in the population at risk and may include cases as well as noncases.24

A total of 2,655 cancers were identified through MCSS and 2,188 were successfully linked to their birth certificates. The subcohort consisted of 8,752 subjects, 15 of which were also cancer cases, but none were WT cases. The entire subcohort was used as a comparison group for the WT cases in order to increase power.

Data was collected from information available on the child's birth certificate. The variables examined for a possible relationship with WT included parental demographic variables: age (<20, 20–24, 25–29, 30–34, ≥35), race (White, non-White), ethnicity (Hispanic, non-Hispanic), education (<high school, high school, >high school), maternal place of birth (U.S. or U.S. territory, foreign country); maternal pregnancy variables: number of prior births (none, 1–2, 3 or more), prior fetal loss (none, 1, 2 or more), interval since last fetal death (none, ≤3 years, >3 years), interval since last live birth (none, ≤3 years, >3 years), prepregnancy or gestational diabetes (yes/no), hypertension (yes/no), anemia (yes/no), tobacco use during pregnancy (yes/no), alcohol use during pregnancy (yes/no), drug use during pregnancy (yes/no), weight gain during pregnancy (≤24 lbs, 25–30 lbs, 31–36 lbs, >36 lbs), Kotelchuck index of prenatal care25 (adequate, less than adequate), intrauterine procedures during pregnancy (yes/no), labor induction (yes/no) and type of delivery (c-section or vaginal birth); and child variables: birth weight (<2,500 g, 2,500–4,000 g, >4,000 g), season of birth (winter, spring, summer fall), gestational age (<37 weeks, ≥37 weeks), size for gestational age26 (small, average, large), Apgar score at 1 min (≤7, >7), Apgar score at 5 min (≤7, >7), congenital abnormalities (yes/no), birth plurality (single birth, multiple birth) and sex (male, female). The availability of variables collected changed over time, as is noted in tables of results. For cases, information on age at diagnosis and laterality of the tumor was also collected from MCSS.

Hazard ratios (HR) and 95% confidence intervals (CI) were calculated from stratified Cox regression models using an asymptotic variance appropriate for case-cohort studies.27 If there were less than 5 cases per cell, no regression model was run since very small numbers could cause parameter estimates to become unstable. Models adjusted only for birth year and sex (initial model) were used for all variables due to the small number of WT cases and since not all variables were collected in all years. A multivariate model was constructed which included those variables with a p-value of less than 0.2 in the initial analysis and was limited to the years in which these variables were collected. Given the exploratory nature of this analysis, no adjustments for multiple comparisons were made. Separate analysis was performed for the subset of patients with unilateral disease since the majority of cases with bilateral tumors involve inherited germline mutations.28 We also conducted separate analyses of cases diagnosed younger than 2years of age and those diagnosed at 2 years of age or older. All analysis was performed using SAS 9.1 (SAS institute, Cary, NC).

Result

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Result
  5. Discussion
  6. References

A total of 158 cases of WT were diagnosed in the state of Minnesota from 1988–2004 in children under the age of 15, 138 of which were successfully matched to a birth record. Linked and nonlinked cases were not different from one another with regard to age at diagnosis, year of diagnosis, birth year or sex (data not shown). Of the WT cases, 50 (36%) were diagnosed <2 years of age and 88 (64%) aged 2 or older (Table I). The 17 (12%) cases that had bilateral tumors were an average of 2.3 years of age at diagnosis, which was lower than the average age of 3.3 years at diagnosis of unilateral cases.

Table I. Demographics and Risk of Wilms Tumor
VariableNo. in subcohort (%)No. of cases (%)HR195% CI
  • 1

    Adjusted for sex and birth year.

  • 2

    Collected in birth years 1989–2004.

Maternal race    
 White7,909 (91.3)128 (93.4)Ref 
 Non-White754 (8.7)9 (6.6)0.620.31–1.23
 Missing891  
Maternal ethnicity2    
 Hispanic160 (3.5)3 (2.9) 
 Non-Hispanic4,475 (96.5)101 (97.1)  
 Missing3497  
Maternal age    
 <20753 (8.6)12 (8.7)0.960.51–1.81
 20–242,052 (23.4)19 (13.8)0.530.30–0.91
 25–293,040 (34.7)52 (37.7)Ref 
 30–342,070 (23.7)44 (31.9)1.080.72–1.63
 ≥35837 (9.6)11 (8.0)0.610.32–1.19
 Missing00  
Maternal education    
 Less than 12 years902 (10.9)10 (7.4)0.770.39–1.54
 12 years3,173 (38.3)45 (33.3)Ref 
 More than 12 years4,211 (50.8)80 (59.3)1.150.79–1.67
 Missing4663  
Birth place of mother    
 U.S. or U.S. territory8,227 (94.1)134 (97.1) 
 Foreign country517 (5.9)4 (2.9)  
 Missing80  
Parental marital status    
 Married6,990 (79.9)117 (84.8)Ref 
 Not married1,754 (20.1)21 (15.2)0.620.38–0.99
 Missing80  
Laterality    
 Unilateral 121 (87.7)  
 Bilateral 17 (12.3)  
Age at diagnosis    
 0–<2 50 (36.2)  
 2–<4 46 (33.3)  
 4–<6 32 (23.2)  
 ≥6 10 (7.3)  
Year of diagnosis    
 1988–1995 59 (42.8)  
 1996–2004 79 (57.2)  

Parental demographic information for cases and the selected subcohort is presented in Table I. There was an association between WT and maternal age with a lower risk for mothers aged 20–24 compared to mothers aged 25–29 (HR = 0.53; 95% CI, 0.30–0.91) and a somewhat lower risk for mothers aged 35 or older compared to mother aged 25–29 (HR = 0.61; 95% CI, 0.32–1.19). Marital status was also found to be associated with WT. Children born to parents who were not married had lower risk than those born to parents who were married (HR = 0.62; 95% CI, 0.38–0.99). Paternal demographics were similar to maternal demographics (data not shown). No other demographic variables showed an association with WT either in initial or multivariate analysis (data not shown). In addition, no reproductive history or pregnancy variables were found to be associated with risk of WT (Table II).

Table II. Maternal Reproductive History, Pregnancy Conditions, Birth Procedures and Risk of Wilms Tumor
VariableNo. in subcohort (%)No. of cases (%)HR195% CI
  • 1

    Adjusted for sex and birth year.

  • 2

    Not collected in 1980.

  • 3

    Collected in birth years 1989–2004.

  • 4

    Prepregnancy or gestational.

  • 5

    Collected in birth years 1992–2004.

  • 6

    Collected in birth years 1980–2004.

  • 7

    Collected in birth years 1976–1979 and 1989–2004.

Number of previous live births    
 None3,449 (40.1)61 (44.5)Ref 
 1–24,253 (49.4)64 (46.7)0.840.59–1.20
 3 or more903 (10.5)12 (8.8)0.760.41–1.42
 Missing1471  
Prior fetal loss    
 None6,713 (78.2)108 (79.4)Ref 
 11,348 (15.7)20 (14.7)0.860.53–1.40
 2 or more518 (6.0)8 (5.9)0.840.41–1.74
 Missing1732  
Interval since last fetal death2    
 No prior fetal loss6,566 (80.3)108 (83.1)Ref 
 ≤3 years891 (10.9)13 (10.0)0.880.49–1.58
 >3 years722 (8.8)9 (6.9)0.700.35–1.39
 Missing5738  
Interval since last live birth    
 No prior live birth3,449 (40.2)61 (45.2)Ref 
 ≤3 years3,335 (38.9)48 (35.6)0.820.56–1.21
 >3 years1,786 (20.8)26 (19.3)0.800.51–1.28
 Missing1823  
Maternal hypertension    
 No7,680 (95.8)122 (94.6)Ref 
 Yes337 (4.2)7 (5.4)1.320.61–2.86
 Missing7359  
Maternal anemia3    
 No4,613 (98.3)103 (99.0) 
 Yes80 (1.7)1 (1.0)  
 Missing2917  
Maternal diabetes45    
 No3,195 (96.7)79 (98.8) 
 Yes110 (3.3)1 (1.3)  
 Missing2156  
Tobacco use during pregnancy3    
 No4,011 (85.1)90 (87.4)Ref 
 Yes702 (14.9)13 (12.6)0.860.48–1.56
 Missing2718  
Alcohol use during pregnancy3    
 No4,587 (97.7)101 (98.1) 
 Yes106 (2.3)2 (1.9)  
 Missing2918  
Drug use during pregnancy5    
 No3,177 (98.6)76 (96.2) 
 Yes45 (1.4)3 (3.8)  
 Missing2987  
Weight gain during pregnancy3    
 ≤24 lbs1,029 (26.7)17 (19.3)Ref 
 25–30 lbs1,295 (33.6)32 (36.4)1.460.80–2.67
 31–36 lbs616 (16.0)16 (18.2)1.630.81–3.26
 >36 labs910 (23.6)23 (26.1)1.530.81–2.90
 Missing1,13423  
Kotelchuck index of prenatal care6    
 Adequate4,749 (67.4)90 (73.8)Ref 
 Less than adequate2,300 (32.6)32 (26.2)0.780.51–1.18
 Missing1,15116  
Intrauterine procedures    
 None8,183 (97.6)127 (97.7) 
 Any203 (2.4)3 (2.3)  
 Missing3668  
Induction of labor7    
 No3,879 (73.0)72 (68.6)Ref 
 Yes1,432 (27.0)33 (31.4)1.130.74–1.73
 Missing2256  
Caesarian section    
 No6,954 (82.4)103 (78.0)Ref 
 Yes1,485 (17.6)29 (22.0)1.280.84–1.95
 Missing3136  

Child demographic and birth data is presented in Table III. High birth weight was associated with increased risk of WT for children weighing over 4,000 grams compared to those weighing between 2,500 and 4,000 grams (HR = 1.54; 95% CI, 0.99–2.40). Size for gestational age was found to be associated with WT. Children large for gestational age appeared to be at increased risk compared to those who were average for gestational age (HR = 1.47; 95% CI, 1.00–2.14). Those with gestational age less than 37weeks had a higher rate of WT compared to those born at 37weeks or after (HR = 1.61; 95% CI, 0.97–2.67). There was also an increased risk of WT in those children with congenital abnormalities compared to those without (HR = 3.20; 95% CI, 1.37–7.47); specific abnormalities reported for cases were malformed genitalia (n = 1), rectal atresia/stenosis (n = 1), cleft lip/palate (n = 2), other musculoskeletal (n = 1), Down syndrome (n = 1), other chromosomal abnormality (n = 1) and other abnormality (n = 1). There was a much wider array of congenital abnormalities in the subcohort, with the majority of abnormalities classified as other musculoskeletal abnormalities (n = 19), heart malformations (n = 10) and malformed genitalia (n = 12). No other variables were found to be associated with WT in either the initial or multivariate analysis (data not shown). However, it is notable that 1 out of 138 (0.7%) children with WT was part of a triplet birth whereas only 10 out of 8,751 (0.1%) in the subcohort were part of a triplet birth.

Table III. Child's Birth Characteristics and Risk of Wilms Tumor
VariableNo. in subcohort (%)No. of cases (%)HR195% CI
  • 1

    Adjusted for sex and birth year.

  • 2

    Imputed gestational age based on date of last menstrual period, if missing used physician's estimate when available.

  • 3

    Collected in birth years 1980–2004.

  • 4

    Collected in birth years 1981–2004.

  • 5

    Collected for birth years 1989–2004.

Child's gender    
 Male4,502 (51.5)69 (50.0)0.980.70–1.38
 Female4,243 (48.5)69 (50.0)Ref 
 Missing70  
Birth weight    
 <2,500 g452 (5.2)9 (6.5)1.190.57–2.46
 2,500–4,000 g7,076 (81.2)103 (74.6)Ref 
 >4,000 g1,184 (13.6)26 (18.8)1.540.99–2.40
 Missing400  
Season of birth    
 Winter2,009 (23.0)38 (27.5)1.470.90–2.40
 Spring2,273 (26.0)35 (25.4)1.120.68–1.86
 Summer2,285 (26.1)29 (21.0)Ref 
 Fall2,185 (25.0)36 (26.1)1.310.80–2.15
 Missing00  
Plurality    
 Singleton birth8,528 (97.4)136 (98.6) 
 Multiple birth224 (2.6)2 (1.4)  
 Missing00  
Gestational age23    
 Less than 37 weeks657 (8.3)18 (13.2)1.610.97–2.67
 37 weeks or more7,251 (91.7)118 (86.8)Ref 
 Missing2922  
Size for gestational age23    
 Small271 (3.4)8 (5.9)1.770.81–3.89
 Average5,710 (72.4)85 (62.5)Ref 
 Large1,905 (24.2)43 (31.6)1.471.00–2.14
 Missing3142  
Any congenital abnormality    
 No8,628 (98.6)131 (94.9)Ref 
 Yes124 (1.4)7 (5.1)3.201.37–7.47
 Missing00  
One minute Apgar score4    
 ≤71,798 (23.4)32 (23.7)1.030.69–1.56
 >75,886 (76.6)103 (76.3)Ref 
 Missing3363  
Five minute Apgar score4    
 ≤7275 (3.6)5 (3.7) 
 >77,396 (96.4)130 (96.3)  
 Missing3493  
Assisted ventilation5    
 None4,549 (98.3)97 (97.0) 
 Any79 (1.7)3 (3.0)  
 Missing35611  

While most associations remained consistent in the subgroup analysis of those less than 2 and those 2 or older at diagnosis a few differed from the overall analysis. Large size for gestational was a significant risk factor for those less than 2 (HR = 1.90; 95% CI, 1.04–3.48), but not for those 2 or older (HR = 1.25; 95% CI, 0.77–2.03). High birth weight was a significant risk factor for those 2 or older (HR =1.78; 95% CI, 1.04–3.04), but not for those less than 2 (HR = 1.17; 95% CI, 0.54–2.54). In addition, there was a significant association between WT and delivery by c-section (HR = 1.86; 95% CI, 1.15–3.01) for children 2 or older and a significant association between WT and birth in the fall compared to birth in the summer for children younger than 2 (HR = 3.46; 95% CI, 1.38–8.67).

Limiting the analysis to only those cases with unilateral disease did not materially alter results (data not shown). Multivariate analysis was similar to the initial model analysis, so only the initial model results are presented.

Discussion

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Result
  5. Discussion
  6. References

Little is known about nonsyndromic risk factors for WT. This study was undertaken to try to elucidate some of these potential risk factors using a combination of data sources. Population based cohort data for childhood cancer, such as was used in this study, is not frequently available in the United States. While we found limited associations in our data, this study augments a sparse literature on risk factors for WT using a robust design.

High birth weight has long been suspected to increase risk for WT. We found a marginally significant association of birth weight >4,000g and of large size for gestational age with WT. Previous case-control studies have shown both increased risk for high birth weight13–15 and no association.10–12 A cohort study in Norway found no association between WT and birth weight.17 Two other studies have compared case series to standard population data. One found a higher than expected number of high birth weight cases diagnosed in the first 2 years of life.16 The second study reported a higher proportion of WT cases than expected above the median birth weight for the United States, even after excluding those with overgrowth syndromes.29 Olshan suggested that high birth weight could be associated with not only congenital abnormalities of overgrowth, but also with changes in expression in the IGF-II gene or other genes that regulate growth found on the short end of chromosome 11 where the WT gene is also located.30 Subsequent studies have provided some support for changes in expression of the IGF-II and H19 genes in children with WT.31, 32 One recent study examined the relationship between WT and size for gestational age and reported an odds ratio very similar to our hazard ratio with a 48% increase in risk for children who were larger for gestational age compared to those who were average for gestational age.33 Our results are thus concordant with previous literature and extant hypotheses.

The subgroup analysis by age at diagnosis produced some seemingly divergent results, with an association seen among younger children for large size for gestational but not high birth weight and vice versa among older children. These differences are most likely due to categorization of continuous variables and small sample sizes within each subgroup since, in a test for homogeneity, these hazard ratios were not found to be substantially different from one another. Both of the significant results indicate that large size at birth as a risk factor for WT and are thus consistent with the main study findings.

WT is known to be associated with specific congenital abnormalities including overgrowth syndromes, abnormalities of the genitourinary tract, aniridia and syndromes of hemihypertrophy.2, 34 There has also been some evidence of an association between other congenital abnormalities and WT.11 Our finding an increased risk for WT in children with any congenital abnormality is thus consistent with previous literature. However, in this study, we could not accurately separate those abnormalities typically associated with WT due to broad groupings of abnormalities and the small number of abnormalities observed in the cases.

In this study, an association was found between maternal age and marital status and WT. It was found that women aged 20–24 were less likely to have a child who developed WT compared to those women aged 25–30. We also found a decreased risk of WT in children with unmarried parents. In subgroup analysis, we found an increased risk for those born in the fall compared to those born in the summer for children less than 2 at diagnosis and an increased risk for delivery by c-section for children aged 2 and older at diagnosis. These associations lack biological justification and are therefore more likely consistent with chance or could represent some biological process not yet identified.

Several strengths and limitations of this dataset should be noted. An important strength of this study was the use of population-based data which combined Minnesota birth records with data from the Minnesota Cancer Registry. The case-cohort design is robust to recall bias and selection bias, which frequently affect studies of childhood cancer which actively recruit participants. However, children who migrated out of Minnesota and subsequently developed childhood cancer were not included in this dataset, nor were cancers that occurred in children residing in Minnesota during 1976–1987 recorded. Since WT is exceedingly rare, neither of these limitations is likely to have biased the results substantially.

Reliability and validity is limited for a number of items collected on birth certificates. A recent review determined that while items such as birth weight, Apgar score and delivery methods are generally reliable, other items such as tobacco use, alcohol use, prenatal care and pregnancy complication are substantially less valid when recorded from birth certificates.35 Similar variability in accuracy of birth certificate data has been noted elsewhere.36, 37 Our findings involved mostly variables with good validity and reliability, such as birth weight and maternal age, and hence are credible.

A final limitation was the relatively small number of cases, which restricted study power and our ability to conduct multivariate analysis. Nevertheless, we had 80% power to detect approximate hazard ratios of 2.99, 1.73 and 1.61 for dichotomous variables with prevalences of 5, 25 and 50%, respectively, in the subcohort. It is worth noting that the National Wilms Tumor Study has assembled data on several thousand cases in North America and has previously reported on birth weight.29, 38 However, the study has limited data on other potential risk factors and lacks a control group. By contrast smaller studies such as ours, which have been the norm, have investigated a broader scope.11, 12, 14, 15, 17, 18, 20

In conclusion, this analysis contributes additional evidence that children with large size at birth have an increased risk of WT. We also corroborated known associations of WT with congenital anomalies. Lastly, within the constraints of our sample size, we have demonstrated no association of other birth factors with WT using a strong methodology.

References

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Result
  5. Discussion
  6. References