No excess of early onset cancer in family members of Wilms tumor patients
Wilms tumor is one of the few pediatric cancers with well-defined familial and genetic components. The authors assessed the risk of early-onset cancers in first- and second-degree relatives of patients enrolled by the National Wilms Tumor Study Group.
Using a stratified sampling scheme that targeted 530 families of patients who were believed a priori to have a genetic contribution to their disease, the authors conducted interviews regarding cancer occurrence in 4258 family members from 296 families of patients with Wilms tumor. Reports of malignant neoplasms that occurred before 55 years of age were confirmed by review of medical records wherever possible. A period of risk was defined for each family member based on calendar time and his or her relationship to the proband.
Ninety-nine cancers were observed, whereas 126.8 were expected by applying standard cancer rates for age and calendar period to the 120,885 person-years at risk. The standardized incidence ratio (SIR) was O-E = 0.78 with 95% confidence interval (CI) of (0.64, 0.95). In subgroup analyses, the highest relative risks were observed for parents of the index case (O/E = 21/13.0 = 1.6, 95% CI = 1.0, 2.5) and for leukemia (O/E = 9/4.9 = 1.9, 95% CI= 0.85,3.5).
The results of this study may provide reassurance to families of children who have had Wilms tumor. Potential sources of bias included the low (56%) rate of participation of targeted families. In general, the biases might have led to the underreporting of some cancers, especially in more distant relatives. The possibility of a slight excess of cancer in parents of Wilms tumor patients could not be excluded. Cancer 2001;92:1606–12. © 2001 American Cancer Society.
Wilms tumor is an embryonal tumor of the kidney primarily seen in children less than 8 years of age. It is thought to arise as a result of abnormal development within the metanephric blastema or primitive kidney. It is the one of the few pediatric cancers with well-described familial and genetic components. Genetic predisposition and/or common family environment may contribute to an increased cancer risk in families. The purpose of this study was to investigate if a higher risk of cancer exists within families of Wilms tumor patients, suggesting an inherited cancer susceptibility.
Certain children with specific anomalies or syndromes are known to be at increased risk of developing Wilms tumor.1, 2 In addition, the National Wilms Tumor Study Group (NWTSG) reports a 1.5 % incidence of familial cases in newly diagnosed patients.3 Four different predisposing genes have been associated with Wilms tumor. WT1 is a tumor suppresser gene located at 11p13 and its mutations or deletions are associated with Wilms tumor, aniridia, genitourinary malformation, mental retardation (WAGR) syndrome, Denys-Drash syndrome, aniridia, and genitourinary anomalies.4–6WT1 is known also to play a role in acute myelogenous leukemia and desmoplastic small round cell sarcomas.7, 8 A second Wilms locus, WT2, is believed to exist at 11p15 and probably is the gene region affected in Wilms tumor patients with hemihypertrophy and Beckwith-Wiedemann syndrome.9, 10
Nephrogenic rests are persistent areas of histology within kidneys that resemble metanephric blastema and are considered to be Wilms tumor precursor lesions. These rests are found in up to 40% of kidneys from children with unilateral Wilms and close to 100% of those with bilateral disease. Two types of nephrogenic rests have been described by Beckwith, perilobar and intralobar.11 Perilobar rests were associated with older age at diagnosis and hemihypertrophy. Intralobar rests were seen in younger patients and associated with aniridia, bilateral disease, and genitourinary abnormalities such as cryptorchidism and hypospadias.
Linkage to genes at 19q and 17q12-21 has been demonstrated in certain cases of familial Wilms tumor.12, 13 The influence of genetic factors in Wilms tumorigenesis is further implied by the difference in racial distribution of Wilms tumor. International data though the 1980s reveals a higher incidence in children of African origin and a lower incidence in Asians.14–16
As differing genetic events have been ascribed to cause one type of cancer, so may one genetic event lead to different types of cancers. Cancers that arise from a constitutional mutation in cancer predisposing genes may produce single cancers in a family or multiple primary cancers in an individual. The dominantly inherited Li-Fraumeni cancer syndrome is the result of the loss of activity of TP53, a tumor suppressor gene.17 A diverse group of tumors have been described in affected family members. Other inherited cancer syndromes, such as hereditary breast cancer syndromes, familial colorectal cancer syndromes, and retinoblastoma, may also be manifested by different malignancies.18–20 Several studies have found Wilms tumor patients to have a significant risk of second malignancies.21–24 Although the methods differed, the relative risks for second malignancy ranged from 5 to 11.21, 23 Although there is a strong association with prior therapy, especially radiation therapy, genetic factors may also be contributory.
The relation between Wilms tumor and other cancers has had little investigation. Hartley et. al. proposed that Wilms might be an uncommon component of the Li-Fraumeni syndrome.26, 26 This finding was not supported by Moutou and associates, who investigated the cancer incidence and mortality of 501 French families who had a child with Wilms tumor.27 A small excess of cancers, which were found when we used the Standardized Mortality Ratio, was not statistically significant.
We hypothesized that there may be an association between Wilms tumor and early (< 55 years of age) onset cancer, a marker for inherited cancer, in first-degree and in certain second-degree relatives. We also planned to identify families with a possible inherited cancer syndrome for future study.
MATERIALS AND METHODS
Study probands and their families were ascertained from patients diagnosed in the United States who were registered with the NWTSG between 1970 and 1992. Most of these patients participated in the NWTSG Long Term Follow-up Study, so their whereabouts were known. Five hundred thirty families were initially selected using a stratified sampling procedure that targeted subgroups of patients who might be at higher risk of having genetically inherited factors because of specific clinical and/or histologic characteristics of their Wilms tumor (Table 1). Institutional review board approval was obtained from the Fred Hutchinson Cancer Research Center where the NWTSG data center is located. Families could be contacted only if the principal investigator at the referring institution so authorized. For families with two or more members registered with the NWTSG, the patient with the earliest date of diagnosis was designated as the proband.
Table 1. Stratified Sampling Subgroups and their Incidence of Cancer
|Beckwith-Wiedemann Syndrome||40||13||6||21 (3)||6|
|Second malignant neoplasm||42||12||8||22 (4)||11|
|Bilateral Wilms tumor||60||15||15||30 (2)||6|
|Intralobar nephrogenic rest||59||18||12||29 (1)||5|
|Perilobar nephrogenic rest||60||16||7||37 (5)||21|
|Family history of Wilms tumor||53||6||7||40 (4)||10|
|Random sample of remaining patients||60||11||17||32 (2)||14|
Families whose location was unknown were tracked using standard methods. Data collection began with a letter explaining the purpose of the study, the information to be gathered, and the right of refusal to participate. Patients 18 years of age or older or the patient's parents were invited to call the NWTSG office and schedule a telephone interview. If the family did not respond to the mail solicitation, then attempts were made to reach the family by telephone.
The telephone interview lasted approximately 50 minutes. Information was obtained regarding the health history of the NWTSG patient, the patient's parents, offspring (if appropriate), siblings, half-siblings, grandparents, aunts, and uncles. Requested data on each family member included the date of birth and death, birth defects, occurrence of any malignancy, and date of diagnosis. If a family member was thought to have had a cancer before age 55 years, permission was obtained to contact the affected family member for corroboration of the malignancy. If the affected family member confirmed a cancer, then we sought a medical release and medical record documentation. If a family had an early-onset cancer, then a sequential sampling method was used, such that the interview was extended to the immediate (nuclear) families of the affected person.28
Early-onset cancers were coded according to the 9th Revision of the International Classification of Diseases (ICD-9). Benign tumors, nonmelanoma skin cancers, and familial Wilms tumors were excluded from analysis. Originally, we had hoped to confirm all reported cancers using medical record documentation. This proved impossible because of factors such as nonresponse from hospitals and state death registries or destroyed records. Thus, some cancers were confirmed by review of collected data by an oncologist (J.F.) and experienced interviewer (J.B.). Details used for confirmation were provided by families and usually included specifics, such as names of chemotherapy agents, which corroborated a cancer diagnosis despite the lack of official documentation.
The expected numbers of cancers in different subgroups were determined by multiplying the observed person-years of observations for family members by the Connecticut Tumor Registry or the Surveillance, Epidemiology, and End Results (SEER) cancer incidence rates specific for age, gender, and calendar year using 5-year intervals.29 The Connecticut registry rates were used for the period 1935–1970, and the SEER rates from 1971 on. Person-years at risk for parents were calculated from the year of the proband's birth until the date of completion of the family's interview (1994–1997), the date of death, or attainment of 55 years of age, whichever came sooner. Person-years for grandparents were calculated from the year of birth of the appropriate parent. For all other family members, person-years were calculated from the date of birth. Standard incidence ratios (SIRs) were calculated as the ratio of observed to expected (O/E) number of cases. Ninety-five percent confidence intervals (CIs) were based on the assumption of a Poisson distribution for the observed numbers. They were virtually identical to confidence intervals obtained using an empirical variance calculation that allowed for possible familial clustering of cancer.
We undertook an exploratory analysis of individual families and searched for high-risk families in two ways. First, the pedigrees for all families with early-onset cancers were reviewed for specific types of cancers that might fit a known cancer syndrome. Second, we compared the observed and expected numbers of cancer cases for each individual family given the ages and number of the family members.
Two hundred ninety-six (56%) of the 530 families selected were ultimately interviewed (Table 1). No contact was initiated on 76 families because either permission was denied by the referring institution or the patient had requested previously to be dropped from the long-term study. Forty-five families refused to participate after being contacted, and 19 families never responded to repeated mailings and telephone messages. Eighty-eight families could not be located and, for 6 families whose location was known, no contact was established before the termination of the study. Of the 296 interviews, 26 were initiated but were unable to be completed in the time span of this study.
The median number of family members was 15, with a mean of 15.5 and a range of 3 to 38. The total number of family members included 2157 females and 2101 males. Informants for the primary interview consisted of 44 NWTSG patients, 213 mothers of NWTSG patients, and 39 others. The informant was from the proband mother's side of the family in 72% of cases. Forty families with familial Wilms tumor were interviewed. Fifteen of these families had 2 NWTSG patients, and 2 of them had 3 NWTSG patients (including the proband).
One hundred five early-onset cancers were confirmed. Six were then excluded from analysis because they occurred either before 1935 (n = 2) or before the birth of the proband (n = 4). Sixty-nine percent were confirmed by medical record. The remaining 31% were confirmed by information gathered by family interviews. Two patients had two pathologically different cancers of the same organ diagnosed at the same time. We elected to count each of these as having only one cancer.
There were 120,885 at-risk patient-years. The expected number of cancers for the entire cohort was 126.8 with 99 confirmed (O/E = 0.78). The most frequent types of cancers were breast, leukemia or lymphoma, lung, and genitourinary tumors (Table 2). We did not see the expected numbers of breast cancers. Breast cancer cases also were broken down by age groups, and there was no excess in very early-onset breast cancer (Table 3). The only cancers seen in excess were ovarian cancers and leukemias. However, none of these differences were statistically significant.
Table 2. Observed and Expected Numbers of Cancers by Site
|Uterine (cervix & corpus)||6|
Table 3. Breast Carcinoma Incidence According to Age Group
|< 30 years||1||0.62|
We confirmed more cancers than expected within the cohort of parents and siblings (Table 4). Among the 21 parental cancers, there were 4 breast cancers, 1 ovarian, 4 melanomas, 3 colon cancers, and 2 testicular cancers. Another parent had had testicular cancer diagnosed before the birth of the proband, so he was not included in the analysis. The number of siblings was inadequate for definitive analysis. Beyond the nuclear family, the number of cancers observed fell below those expected. If the data in Table 4 is collapsed to compare cancer incidence in the nuclear family versus distant relatives, there is a statistical difference of P = 0.001 (Table 5).
Table 4. Relationship of Cancer-Affected Relative to Proband
Table 5. Ascertainment in Distant Relatives
We found fewer malignancies than would have been expected within the overall cohort of Wilms tumor families. Two cancers that occurred prior to the initiation of the Connecticut registry database in 1935 were not included into the analysis as there were no comparable reference rates and they made little difference in the observed to expected ratio.
The SIR for parents was elevated at 1.62 but with a 95% confidence interval of 1.00–2.48. While this is intriguing, a prior study from Denmark showed no increased risk of malignancy in the parents of children with cancer. In fact, fathers of Wilms tumor patients were found to have a lower than expected incidence of cancers.30
The Moutou study asked a similar question and reached a similar overall conclusion, mainly the lack of a significant global excess of cancer in the families of Wilms tumor patients.27 They did report an excess of bone (5) and brain cancers (5). Our study, which involved about half as many families, identified only two brain tumors and one bone cancer and so lacked sufficient power to analyze these specific disease sites.
Several potential sources of bias existed that might have led to an erroneous number of reported cancers within a family. In most instances these biases would have led to underreporting of cancers. First, we were able to conduct interviews on only 56% of the initially selected families. It is difficult to know whether these families were representative of the whole group. The percentage of families who refused to participate, actively or passively, ranged from 15% in the familial Wilms group to 36% in the aniridia group. The familial group had been tracked and studied extensively, so that more releases and addresses were available at the start of this study.
Second, we expected, and probably found, underreporting, especially in distant relatives (Table 5). Some informants with whom we spoke with had little to no contact with the relatives that were of interest to the current study. One cannot confirm a cancer that never is reported by a family member. Although the 95% CIs for the SIR included the value 1.0, we did confirm more cancers than expected within the cohort of parents and siblings. This is the group in which we expected to find the most accurate information. There is data suggesting that patients with genitourinary anomalies, or the WAGR, or the Denys-Drasch syndromes may carry de novo mutations.31 Thus, other family members would not be carrying a mutation, and no increased familial risk would be seen.
Third, confirming reported cancers is a very difficult task in the United States. Unlike many European countries, there is no national database of cancer diagnoses. Obtaining consent for release of medical records from an affected individual proved frustrating, if not impossible, at times. Many medical records had been destroyed or were not available. Cancers were reported that might have been legitimate early-onset cancers, but records did not exist, or remaining family members did not have enough information to allow us to confirm these cancers. Of the 26 partial interviews, most were incomplete because the mother no longer had contact with the father's family, or secondary interviews on affected grandparents were impossible because the great-grandparents were deceased.
Both the small sample size and under reporting might have lead to insufficient power to detect a statistical increase of early-onset cancers in these families. However, if we assume that the remaining 44% that were not interviewed had similar ages, gender, and family numbers as the reported group, then we can demonstrate that even underreporting probably did not change the results of the study. If the noncontacted group had a 50% higher incidence of early-onset cancers, then the SIR would be 0.95 with a 95% CI (0.83, 1.09). If the noncontacted group had even a 100% higher incidence, then the SIR still would be only 1.13 with a 95% CI of 0.99–1.27.
Finally, a gender reporting-bias probably was introduced as the majority of interviewees were mothers of the probands. It is likely that some cancers from the father's side were missed because of the mother's lack of knowledge. Forty cases were reported from the father's side, with 61.4 expected. Fifty-seven were reported in the maternal family, with 64.1 expected. Two were reported from both (cancer in a sibling), with 1.4 expected.
One of our goals was to search for families that were suggestive of a family cancer syndrome. Some of the cancers discussed below were not considered confirmed by our standards. Other cancers mentioned were confirmed but were not included in our analysis because they would not have been discovered because of limits imposed by the rules for sequential sampling. Two families (Pedigrees A and B) (Fig. 1) had a second child with a malignancy.32 Family A had another child with a brain tumor but no other family members with reported cancers. Family B had three additional confirmed cancers: the sibling with Hodgkin disease, a maternal uncle with childhood leukemia, and a maternal grandfather with prostate cancer. Although Family B did not fit into a defined, inherited-cancer syndrome, it appeared to have an inherited cancer propensity on the maternal side.
Twenty probands had a parent who also had a cancer; in one case, the proband's mother and father both had cancer. Of these 20 probands, there were 4 families (C, D, E, and F) in which a third cancer was reported on the same side of the family. Families C and D appeared to possibly have an inherited breast-cancer syndrome. The proband's father in Family C had a carcinoma of the small intestine. His mother and grandmother had breast cancer. Colon cancer is not uncommonly seen as part of the BRCA1 syndrome. Perhaps other intestinal cancers may occur as the result of the loss of this tumor suppressor. Family D also revealed breast cancer but on the maternal side of the family. Family E was suspicious for an inherited colon-cancer syndrome. Family F had a mother with thyroid cancer and a maternal uncle who had leukemia.
In addition, there are other families that had three to four generations of reported cancers but did not meet criteria for any cancer-predisposing gene. In one of these families, the father of the proband had testicular cancer. The paternal grandmother was diagnosed at age 53 years with colon cancer, and there was a maternal uncle with leukemia. One other father also had testicular cancer. Michael DeBaun, M.D., recently completed a study of Beckwith-Wiedemann patients and found an excess of testicular cancers in family members (unpublished data). The family members he interviewed were a different group than those we contacted.
These data from the current study may be reassuring to families with children who have had Wilms tumors. These data also do raise the possibility of a higher than expected incidence of early-onset cancers in parents. This supports the concept of an inherited oncogene that can manifest as different cancers in different individuals. The risk in siblings is quite difficult to assess, as they constitute such a small and young group. There remains a risk that this is a negative study because of lack of statistical power secondary to low participation and bias that led to underreporting. The most recent Wilms tumor studies in the United States have focused on tumor biology. Future studies should incorporate the ability to prospectively follow family members for early-onset cancers.
The authors thank the Principal Investigators and their staffs from the Children's Cancer Group and the Pediatric Oncology Group, as well as the National Wilms Tumor Study Data and Statistical Center staff for their assistance with this project. Additional thanks to Ellen Wijsman, Ph.D. for review of the article.