Although hepatoblastoma is a very rare childhood cancer, its incidence appears to be rising, especially among children with very low birth weight. With the exception of documented correlations with certain congenital anomalies, the etiology of hepatoblastoma remains largely unknown.
Using California's population-based cancer registry, the authors identified 113 children ages birth–4 years with hepatoblastoma who were diagnosed between 1988 and 1997. Ninety-nine of those 113 children (88%) were matched to a California birth certificate, and randomly selected controls from the same birth certificate files were matched to cases (4:1) according to the month and year of birth and gender. Odds ratios (OR) and 95% confidence intervals (95% CI) were estimated using conditional logistic regression analyses.
A strikingly elevated risk of hepatoblastoma was found in children who were born with very low birth weight (< 1500 g; OR, 50.57; 95% CI, 6.59–387.97). A plot of the distribution by birth weight showed interesting peaks at birth weights < 1000 g and 3000–3499 g among cases. Children who weighed < 1000 g showed a statistically significant, linear trend toward being diagnosed at an older age (P = 0.036), which seemed to be explained in part by gestational age.
Hepatoblastoma is a rare malignant childhood neoplasm. Though it comprises less than 1% of all childhood cancers in the U.S., it is the most common type of hepatic malignancy in children, followed by hepatocellular carcinoma. It usually presents during the first 5 years of life.1 An analysis of Japanese cancer registry data revealed an increasing trend in hepatoblastoma incidence among children of very low birth weight.2 This same trend subsequently was reported in the U.S..3–5 A study that used data collected through the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) Program found that, during 1973–1992, hepatoblastoma incidence increased among U.S. children age < 5 years, coinciding with a period of improved survival rates for infants with low birth weight.6
Despite the rising incidence of hepatoblastoma, the rarity of the disease has made population-based epidemiologic studies extremely difficult, hindering researchers' ability to examine its etiology efficiently. Certain congenital anomalies, including Beckwith–Weidemann syndrome, hemihypertrophy, trisomy 18 syndrome, and familial adenomatosis polypi, have been associated with increased hepatoblastoma risk.7–10 However, these conditions account for only a small proportion of cases. Isolated case reports of sporadically occurring hepatoblastoma suggest associations with oral contraceptive use during pregnancy,11 fetal alcohol syndrome,12 hormone treatment for sterility,13 and maternal liver transplantation combined with immunosuppressive treatment during pregnancy.14 In a case–control study of hepatoblastoma risk factors, the Children's Cancer Group reported statistically significant increased risks associated with certain parental occupational exposures, such as metals, petroleum products, paints, and pigments.15 Recently, in a study of parental smoking and childhood cancers, the United Kingdom Childhood Cancer Study (UKCCS) found elevated hepatoblastoma risks for children of mothers who smoked before conception and for children whose parents both smoked.16 Prenatal exposure to cigarette smoke is an established risk factor for low birth weight.17, 18 The UKCCS findings, together with the growing evidence of a correlation between low birth weight and hepatoblastoma, have led some investigators to speculate that birth weight may confound the observed association between parental smoking and hepatoblastoma risk.19 In a reanalysis by Pang et al. that controlled for the effects of birth weight, parental smoking remained significantly associated with hepatoblastoma, although that study was based on a small sample size.20 The question of whether cigarette smoke exposure acts as an independent risk factor or has an indirect effect, by way of low birth weight, warrants further investigation.
Reports associating prematurity with hepatoblastoma2–5, 21, 22 have moved investigators to suspect that intensive care treatments used in recent years on infants with very low birth weight may be involved. It is possible that the immature kidneys and livers of low-birth-weight infants render them vulnerable to the damaging effects of treatments commonly administered to such neonates.23 One recent case–control study reported that infants who had extremely low birth weight (< 1000 g) who were diagnosed with hepatoblastoma had received relatively longer durations of oxygen therapy and furosemide use compared with infants who had extremely low birth weight without hepatoblastoma.24 We took advantage of California's large, well established, population-based cancer registry data and birth certificate files to conduct a case–control study further examining the correlation between hepatoblastoma incidence and a variety of potential risk factors.
MATERIALS AND METHODS
Using California's statewide cancer registry (the California Cancer Registry), we identified 113 hepatoblastoma cases in children ages 0–4 years who were diagnosed between 1988 and 1997. Using probabilistic record linkage,25, 26 we located California birth certificates for 99 children in the case group (88%; birth years 1983–1997). From the same birth certificate files, we randomly selected four children for the control group for each child in the case group, matched according to month and year of birth and to gender. To be consistent with our case-selection criteria, we chose controls whose mothers were California residents at the time of delivery. We then cross checked the birth certificates of our control children against the California Birth Cohort files, which link birth and death records for infants in the first year of life. Controls who had died in infancy were replaced if their death occurred at an age younger than the corresponding case's age at diagnosis. We later dropped one child from the control group, because we determined that the mother's address at delivery was outside California. Because this was a records-based study, we were not required to obtain informed consent from the patients studied; however, our use of human subject data was reviewed by the California Health and Human Services Agency, Committee for the Protection of Human Subjects and was in compliance with their ethical standards as well as with the U.S. Code of Federal Regulations (Title 45, Part 46 on the Protection of Human Subjects).
Birth Certificate Information
California birth certificates served as our main data source, providing demographic information, pregnancy history, and childrens' birth characteristics. Available birth certificate information on childrens' parents included age, race, Hispanic ethnicity, and place of birth. California began collecting education level for each parent on birth certificates starting in 1989; for children who were born during or after that year, we constructed a household-education variable based on the highest level of education attained by either parent. Available information on maternal pregnancy history included numbers of pregnancies and live births, pregnancy losses (spontaneous abortions and stillbirths), and time since last live birth. The birth certificates also provided information about prenatal care and delivery method (vaginal or cesarean) for each case and control child. Information available for each child included date of birth, place of birth, gender, race/ethnicity, birth weight, gestational age, and whether the birth was single or multiple (twins, triplets, etc.).
We used conditional logistic regression models in a matched analysis to calculate univariate odds ratios (OR) and 95% confidence intervals (95% CI). All variables that were statistically significant in the univariate analyses were included in a multivariate, conditional logistic regression model. We used the Cochran–Armitage test for trend to evaluate the relation between birth weight and age at diagnosis for hepatoblastoma cases. We performed all statistical analyses using SAS software (version 8; SAS Institute Inc., Cary, NC).27
Our analyses included 99 hepatoblastoma cases and 395 individually matched controls. Table 1 shows results from our univariate analyses examining the association between various birth certificate characteristics and hepatoblastoma. Hepatoblastoma incidence appeared slightly higher in males than in females, with a ratio of 1:1.25. Incidence was highest among infants (birth–12 months of age), with 10% of the diagnoses occurring during the first month of life. We also observed a trend toward decreasing incidence during subsequent years of life. Cases and controls appeared significantly different for some infant characteristics. Children with a race/ethnicity designation of “other,” which included Asians, Pacific Islanders, and Native Americans, had a significantly reduced risk of developing hepatoblastoma compared with non-Hispanic White infants (OR, 0.36; 95% CI, 0.15–0.90). We found a strikingly elevated hepatoblastoma risk in children who were born with very low birth weights (< 1500 g; OR, 50.57; 95% CI, 6.59–387.97), whereas children with high birth weights (≥ 4000 g) showed no statistically significant risk. Among infants with very low birth weights, 10 of 13 children (77%) weighed < 1000 g at birth. A plot of the distribution of children by birth-weight categories (Fig. 1) shows an interesting peak in birth weights < 1000 g for cases but not controls. Among hepatoblastoma cases, those who weighed < 1000 g at birth showed a statistically significant, linear trend toward being diagnosed at a later age (P = 0.036); whereas the group with normal birth weight displayed no recognizable trend (Fig. 2). Because children with low birth weight often are premature, children with a gestational age < 37 weeks also had an elevated OR compared with children who were born between 37 weeks and 41 weeks of gestation (OR, 2.56; 95% CI, 1.35–4.84). All children with a birth weight < 1500 g were premature, with a gestational age < 37 weeks. Children who were part of a multiple birth had a nearly 3-fold increased risk of hepatoblastoma at borderline statistical significance (OR, 2.74; 95% CI, 0.94–7.96). The six children in the case group who were born as part of multiple births were unrelated. None of the characteristics we examined related to pregnancy and delivery, the child's parents, or maternal pregnancy history showed evidence of a statistically significant association.
Table 1. Univariate Analysis of Birth Certificate Characteristics and Hepatoblastoma in California Children, 1987–1997
Table 2 shows the results of our multivariate, conditional logistic regression model, which included all statistically significant variables from the univariate analyses. Very low birth weight was the only variable that remained statistically significant, with a slightly lower but still remarkably large point estimate (OR, 40.80; 95% CI, 4.21–395.50). After controlling for birth weight, the effects we saw initially for young gestational age were no longer evident (OR, 0.84; 95% CI, 0.28–2.21). Point estimates for a child's race/ethnicity remained consistent with those in our univariate model. We excluded the multiple-births variable from our final model, because only 1 control with a birth weight < 1500 g was part of a multiple birth, thus generating unstable estimates. Inclusion of this variable into the model yielded an unstable but similarly elevated estimate for the very-low-birth-weight category, whereas estimates for the other variables remained similar (data not shown). In a multivariate model, which included a multiple-births term and combined the two lowest birth-weight categories (< 1500 g and 1500–2499 g), children who weighed < 2500 g at birth also appeared to have a significantly increased hepatoblastoma risk compared with children who had normal birth weights (OR, 4.54; 95% CI, 1.55–13.33; data not shown). The point estimates did not change dramatically when we excluded the multiple-births term from our multivariate model. This suggests that being a twin or a triplet does not markedly influence the association between low birth weight and hepatoblastoma.
Table 2. Multivariate Analysis of Select Birth Certificate Characteristics in Relation to Hepatoblastoma in California Children, 1987–1997
This case–control study offered a unique opportunity to examine the correlation between various birth characteristics and hepatoblastoma risk using a relatively large, population-based sample of individuals. We saw a dramatic increase in hepatoblastoma risk among children who weighed < 1500 g at birth compared with other children. This association also has been observed in Japan. Analyzing 1985–1993 data from the Japanese Children's Cancer Registry, Ikeda and colleagues observed a trend toward increasing percentages of hepatoblastoma among children with birth weights < 1500 g; this trend was not observed for other major childhood cancers.2 A closer look at the data showed that the trend was driven mainly by an increase in the number of hepatoblastoma patients with extremely low birth weights (< 1000 g). In the current study, a breakdown of patients by birth-weight categories revealed that > 75% of children with very low birth weights weighed < 1000 g. However, these data did not support the finding of an increasing trend in hepatoblastoma incidence among children with very low birth weight by year of diagnosis (data not shown).
In another recent study that compared all live births in Japan with hepatoblastoma cases identified by Japan's cancer registry, Tanimura and colleagues reported significantly elevated relative risks of hepatoblastoma among children with low birth weight.21 Examining the correlation between low birth weight and age at diagnosis, those authors found no statistically significant association but noted that the minimum age for hepatoblastoma diagnoses tended to be later for children with low birth weight than for children with normal and high birth weights. We also observed later ages at diagnosis for children with very low birth weight compared with other children. However, when we examined this correlation using postconception age (time from conception to diagnosis) to account for differences in gestational age, the trend was weakened. Differences in age at diagnosis between children with normal and very low birth weights suggest differing etiologic mechanisms in the development of hepatoblastoma. For example, by activating enhancing growth factors, cellular growth abnormalities may play a more pronounced role in children with heavier birth weights. Conversely, exposures that increase oxidative damage may operate more powerfully in premature children. Because of their low body weight or small body surface area, such extremely small infants receive relatively more exposure to parental nutrition, drugs, or radiation from chest and abdominal radiographs.
The SEER data have shown that hepatoblastoma incidence is highest among children age < 4 years.1 Between 1988 and 1997, the California Cancer Registry identified 121 hepatoblastoma cases in children age < 15 years. Of those, 113 cases (> 90%) occurred in children age < 5 years. Approximately 50% of the cases included in the current study were in children age < 1 year at diagnosis, and 10% of these occurred during the first month of life. These observations demonstrate that hepatoblastoma is a disease of early childhood with an etiology likely related to risk factors that occur before conception, during gestation, and/or soon after birth.
Because they used homogenous study populations and small sample sizes, previous studies could not evaluate risk related to a child's race/ethnicity. Our analyses, which included a large sample of California's ethnically diverse population, suggest that there is little difference in risk by racial/ethnic group, with the exception of the lower risk seen in the small heterogeneous group of children identified as Asians, Pacific Islanders, and Native Americans.
If a target population could be refined by identifying infants with very low birth weight, then hepatoblastoma screening could be implemented readily. Almost all true hepatoblastoma cases carry increased α-fetoprotein (AFP) levels at diagnosis or fail to exhibit the normal decline in AFP level after birth. Thus, AFP monitoring may be used to detect cases early (Stages I or II), when surgery and minimal chemotherapy treatments offer excellent survival rates (> 90%).
Our nonintrusive data-collection and case–control selection methods limited us to information available only from California birth certificates and from the California Cancer Registry. One disadvantage of this method is that we lacked potentially important information on personal behavior and environmental risk factors. In addition, birth certificate data sometimes are inaccurate or incomplete. Examining the percentage of missing birth certificate data for each factor of interest, however, we noted at least 99% completeness for all data elements except father's age, gestational age, and time since last live birth. Birth certificate data are most likely to be reliable for some factors that are recorded at the time of birth, such as birth weight and race/ethnicity. A recent validity analysis of race and Hispanic ethnicity information from California birth certificates found that the information was a valid measure for all groups except Native Americans.28
To our knowledge, the current study is the largest population-based, case–control hepatoblastoma study conducted to date. Because this disease is rare, our study had a small sample size, which may have limited our power to evaluate some risk associations. We designed our 4:1 matching scheme to optimize power for examining associations between hepatoblastoma and the birth characteristics of interest. Our study's strengths lie in its population-based patient ascertainment, reduced selection and information biases, and case–control study design. Compared with most previous studies, which relied on small case series, our study's case–control design enabled us to evaluate several etiologically relevant risk factors. We drew our cases from the statewide, population-based cancer registry, which has an estimated 99% ascertainment completeness.29 Although we limited our cases to children who were both diagnosed and born in California, most likely excluding highly mobile patients, our selection included a high proportion of all California cases diagnosed within our study time frame. We eliminated participation bias by selecting controls randomly from the California birth files, rather than using volunteer recruits. The records-based nature of our study minimized the potential information biases common to many epidemiologic studies that use questionnaires or interviews to gather data.
The results of the current study confirm previously reported findings of an increased hepatoblastoma risk among children with low birth weight, and they suggest that the etiology may differ between children with very low and normal birth weights. Hepatoblastoma, like most cancers, however, appears multifactoral in etiology, possibly involving a combination of risk factors, including predisposing conditions, genetic susceptibility alleles, prenatal environmental exposures, and neonatal exposures related to treatment of infants with very low birth weight. A recent report by Pakakasama and colleagues30 provides one biologic clue: There is a reduced risk of hepatoblastoma in children who have the polymorphic A allele within the myeloperoxidase (MPO) gene's promoter region, resulting in decreased expression of the neutrophilic enzyme. We would expect children with such an allelic expression to have less free oxygen radical production. A similar finding has been observed in studies of MPO allelic expression in patients with lung carcinoma.31 Epidemiologic studies examining these risk associations are limited and continue to challenge researchers, because hepatoblastoma is so rare. A larger study that includes biospecimen analyses will be required to examine the relations between hepatoblastoma risk, environmental exposures, and relevant metabolic polymorphisms.
The authors thank the staff of the California Cancer Registry and the staff of the Office of Vital Records. They also thank Eric Elkin and Susan Hurley for conducting the record linkages. Theresa Saunders assisted with article preparation.