Curative but potentially mutagenic cancer therapy might lead to untoward disorders and increased hospitalization among the offspring of childhood cancer survivors. Hospitalizations in childhood were evaluated in a population-based cohort of 1,920 offspring of 3,963 childhood cancer survivors, 6,394 offspring of 5,657 siblings and 9,594 population-based comparisons. The Danish Cancer Registry, Central Population Register and National Hospital Register were used to identify study subjects and hospitalizations. The probability for children in the offspring cohorts of being hospitalized before a given age was estimated using the Kaplan–Meier method. Hospitalization rate ratios (HRRs) were calculated using a Cox proportional hazards model with population comparisons as referent. Little differences in hospitalization histories were seen among offspring in the 3 cohorts. HRRs of overall hospitalization was 1.05 (95% CI, 0.98–1.12) for offspring of survivors and 1.01 (95% CI, 0.97–1.05) for offspring of siblings, neither of which was significantly different from that of population comparisons. No significant associations were seen for most of the main diagnostic groups of diseases including infections and perinatal disorders. A 6-fold excess risk of hospitalization for malignant tumors in survivors' offspring, however, could largely be explained by hereditary cancer syndromes, and part of the 2-fold excess hospitalization for benign tumors might similarly be explained by an underlying genetic susceptibility or by increased surveillance of children born to survivors. Assuming that hospitalization is an indicator of multifactorial genetic disease, the findings provide further reassurance that cancer therapies do not confer a high risk of such conditions in offspring born after treatments.
Modern treatments for cancer include high doses of ionizing radiation and combinations of high-dose systemic chemotherapy. Such therapies have the potential to produce germ-cell mutations leading to genetic disease in the next generation. The continued rise in survival rates for childhood cancer, now about 80%,1, 2 and the ability for survivors to have children of their own3 have highlighted the importance of evaluating the impact of cancer therapy on fertility, pregnancy and health of the growing number of children of cancer survivors.
Previous studies of trans-generational effects among offspring of childhood cancer survivors have primarily considered outcomes with a high genetic component such as chromosomal abnormalities, single gene disorders and certain congenital malformations such as neural tube defects and cleft lips/palates.4–6 Conceivably, any induced transmissible mutations might lead to functional disturbances affecting different organs and systems such as the immune system with increased risks of infections or immunological disorders, the endocrine system resulting in diabetes mellitus, thyroid disease or more rare metabolic disorders, or neurologic disorders such as mental retardation or depression. Multifactorial diseases might also be the result of the joint action of genetic and environmental factors, and result in common congenital abnormalities such as neural tube defects, cardiovascular malformation and cleft lips/palates, i.e., abnormalities that are not always readily explained on the basis of simple Mendelian patterns of inheritance.7
In a population-based study of the children of survivors and adolescent cancer in Denmark, we address whether successful and potentially mutagenic cancer therapy might lead to untoward disorders as measured by hospitalizations for diseases of childhood with attention given to infections as an indicator of a more general impairment of health in the next generation.
Material and Methods
From the Danish Cancer Registry, 3,963 survivors were identified who were diagnosed with cancer at age < 20 years between 1950 and 1996 and who survived until the onset of fertility (age 15 years). Patients had to be born 1950–1984 and alive on April 1, 1968, when the Central Population Register (CPR) was established with personal identification numbers for all citizens that permit linkage among various health registers. The cancer registry records contain information on the date of diagnosis, type of cancer, whether radiotherapy was given and the personal identification number of the patient.8
The CPR also contains up to date information on vital status, migration and first-degree relatives. Accordingly, we identified 5,657 cancer-free siblings of the cancer patients, using the same inclusion criteria. Only survivors and their siblings born 1950 or later were included, as sibling identification before 1950 was incomplete. A search in the CPR also identified all live born offspring of survivors and their siblings born between January 1, 1977 [start of the Danish National Hospital Register (NHR), from which information on hospitalizations was available as described later] and December 31, 2003. In addition, a population comparison group of children were randomly selected from the CPR matched on birth month and year of survivors' children in an approximate 5 to 1 selection ratio.
Hospitalization among offspring
Linking our study rosters to the NHR which records information on nearly 99% of all admissions to nonpsychiatric hospitals in Denmark,9 we identified dates and reasons for any hospitalizations of the offspring of survivors and offspring of the 2 comparison groups from 1977 through 2003. Each admission record includes the personal identification number, the date of admission and discharge, and up to 20 discharge diagnoses coded according to a Danish version of ICD-8 until 199310 and thereafter to ICD-10.11 In this study, diseases requiring hospitalization were included and diagnoses from outpatient and emergency room patient visits (available since 1995) were excluded.
Discharge diagnoses were grouped into 16 main diagnostic categories defined by the ICD-8 and 10-classification. Because most Danish children are born in hospitals, all diagnostic codes related to birth such as vital status, singleton/multiple pregnancy and neonatal biometric parameters were excluded. All selected main diagnostic groups accord with the original disease classification system except for the group of infectious diseases, which consists of the original main group entitled “Certain infectious and parasitic diseases” (chapter I in the ICD-classification) as well as all diagnostic codes from other main groups including site-specific infections (main groups VI-XIV); e.g., pneumonia originally classified within “Diseases of the respiratory system” (X) was reclassified into the group of infectious diseases. Perinatal infections, however, were kept within “Certain conditions originating in the perinatal period” (XVI). The diagnostic group “Pregnancy, childbirth and the puerperium” (XV) was not of relevance for hospitalizations in childhood.
Radiation dose characterization
For survivors treated with radiotherapy, an estimate was made of dose to the ovary, uterus, testis and pituitary gland. Four categories of radiation dose were developed (low, low or medium, medium or high and high). Dose categories were assigned to these organs by a highly experienced medical physics team (Stovall et al.12) based on the type and topography of the childhood cancer being treated, knowledge of treatment procedures for specific tumors during the study period and assumptions of the gonadal proximity to the radiation field (for same approach6, 13).
The categories “low or medium” and “medium or high” were used for grouping of cases when a more accurate organ exposure could not be determined without additional treatment detail. For primary cancer treatment extending below the diaphragm, dose to ovary and uterus was estimated to be medium to high with a dose ranging between 100 and 4,000 cGy and the dose to testis between 20 and 2,500 cGy. For primary cancer treatment above the diaphragm, including treatment for retinoblastoma or brain tumors, the dose to ovary and uterus was between 1 and 100 cGy and the dose to testis less than 20 cGy. Dose categories were defined a priori based on the anticipated adverse effect on gonads; i.e., low radiation doses with expected low or no adverse effect on gonads and germ-cells and high doses which might lead to serious adverse effect on these organs (germ-cell damage, uterine damage). The dose to the pituitary gland was estimated to be high during radiation for brain tumors and leukemia treated with cranial irradiation, typically between 5 and 50 Gy, but low for tumors located below the diaphragm, including treatment for Wilms' and gonadal tumors, ranging from 0.01 to 0.1 Gy.
The probability of being hospitalized before a given age in the three offspring cohorts was estimated using the Kaplan–Meier method. Cox proportional hazards model was used to estimate hospitalization rate ratios (HRRs) in offspring of survivors compared to offspring in the 2 comparison groups taking covariates into consideration. Children were followed from birth until age at first hospitalization using the date of admission, end of age 14, emigration, death, or end of follow-up (December 31, 2003), whichever occurred first. HRRs with corresponding 95% confidence intervals were calculated for overall hospitalization and for hospitalization with a diagnosis within the selected main diagnostic groups, with population comparisons as referent. Hospitalizations might vary over time. Therefore, birth year of offspring was included as a time-dependent variable modeled as a linear spline with knots at 1985 and 1995. As the age of both the mother and the father might influence on the threshold of hospitalization, analyses were adjusted for parental age at time of birth linearly. All analyses were stratified by sex of offspring; i.e., allowing for separate underlying hospitalization intensities in boys and girls. Separate risk estimates were calculated according to characteristics of the survivor; i.e., sex, type of childhood cancer (12 categories),14 age at diagnosis (≤1, 2–4, 5–9, 10–14, 15–19), year of diagnosis (5-year calendar periods), radiotherapy (yes/no) and estimated radiation dose to the gonads (ovary and testes). In a subanalysis comparing offspring of survivors with offspring of siblings, we were able to adjust for birth order, which also might influence on the threshold of hospitalization. Further analyses were conducted for the diagnostic group of infectious diseases by use of the Prentice, Williams and Peterson, total time method (PWP-CP)15; i.e., an extension of survival models based on the Cox proportional hazards approach taking into consideration that hospitalizations are recurrent events.
A search of the CPR for offspring born between 1977 and 2003 resulted in identifying 1,920 live-born offspring of the 3,963 survivors (987 boys and 933 girls after exclusion of 15 children born before or up to 9 months after their parent's cancer diagnosis), 6,394 offspring (3,293 boys and 3,101 girls) of their 5,657 siblings and a population comparison group of 9,594 offspring (4,974 boys and 4,620 girls). Table 1 shows the characteristics of the 1,066 survivors who became parents. In the same calendar period, a total of 33,820 discharge diagnoses were registered in the NHR for all subjects before age 15 with a median follow-up for hospitalizations on 9.6 years (range 0–<15) in all offspring cohorts.
Table 1. Characteristics of 1,066 survivor parents
The probability for offspring of survivors of being hospitalized before a given age in childhood—overall and for selected diagnostic groups of diseases (infections and respiratory diseases shown)—was remarkably close to that in the comparison groups (Fig. 1). A little more than 60% of offspring of survivors were hospitalized before age 15 with similar proportions in the two comparison cohorts. Cumulative hospitalization rates from birth through age 14 for all main diagnostic groups and the most common discharge diagnoses within these main groups are presented in Table 2. With the population comparisons as referent, the adjusted HRRs of overall hospitalization was 1.05 (95% CI, 0.98–1.12) for offspring of survivors and 1.01 (95% CI, 0.97–1.05) for offspring of siblings, neither of which was significantly different from that of offspring population comparisons. No significant differences were seen for most of the main diagnostic groups of diseases between offspring of survivors and offspring of siblings and the population comparison group. Offspring of survivors were not hospitalized more for infectious diseases in general (HRR = 1.0 and a similar cumulative hospitalization rate of nearly 30% in all cohorts) or for common infections such as diarrhoea and acute gastroenteritis, otitis, appendicitis, or the most common respiratory infections (Table 3). A more refined statistical method conducted for overall infections which took into consideration that hospitalizations are recurrent events provided similar results (not shown). Further adjustment for birth order conducted in a subanalysis did not change the risk estimates. Furthermore, no significant differences were seen for the most common perinatal disorders such as prematurity, asphyxia and hyperbilirubinemia (Table 4).
Table 2. Cumulative hospitalization rates and adjusted hospitalization rate ratios (HRRs) for overall hospitalization and for main diagnostic groups of diseases in childhood in 1,920 offspring of childhood cancer survivors, 6,394 offspring of siblings and a population comparison group of 9,594 offspring (referent)
Table 3. Adjusted hospitalization rate ratios (HRRs) for common types of infections in childhood in offspring of childhood cancer survivors, offspring of siblings and a population comparison group (referent)
Table 4. Adjusted hospitalization rate ratios (HRRs) for common types of perinatal disorders in offspring of childhood cancer survivors, offspring of siblings and a population comparison group (referent)
Offspring of survivors, however, had a 6-fold significantly increased risk of being hospitalized for a malignant tumor based on 20 observed cases (Table 2). In 9 families, tumor combinations in the cancer survivor and his or her child/children were compatible with well known hereditary cancer syndromes and most likely not a result of treatment-induced germinal mutations. With the exclusion of 8 families with retinoblastoma and 1 family suggestive of multiple endocrine neoplasia 2A, the HRR for malignant tumor was reduced from 5.7 (3.0–10.8) to 2.0 (7 observed cases, 95% CI 0.8–4.8) when compared to population comparisons.
Also a 2-fold significantly increased risk of being hospitalized for a benign tumor was observed among offspring of survivors (22 observed cases). Two survivors of medullary thyroid carcinoma both had 2 children diagnosed with a neoplasm of endocrine glands not further specified which probably were benign thyroid tumors and 1 survivor of retinoblastoma had a child with a benign neoplasm of the optic nerve, suggesting a genetic susceptibility or an increased surveillance of offspring of cancer survivors. Excluding these families the risk was reduced from 2.0 to 1.5 (17 observed cases, 95% CI 0.9–2.7).
The overall hospitalization rate ratio was slightly higher in boys (HRR = 1.12, 95% CI, 1.03–1.23) than in girls born of survivors (HRR = 0.96, 95% CI, 0.87–1.06) based on population comparisons primarily due to a nonsignificantly increased risk among boys of being hospitalized with an infection (HRR = 1.09, 95% CI, 0.95–1.24), a perinatal disorder (HRR = 1.12, 95% CI, 0.97–1.30) or a congenital malformation (HRR = 1.23, 95% CI, 0.96–1.59) (not shown). A similar pattern between the 2 sexes, however, was also seen in offspring of siblings (HRR = 1.06, 95% CI, 1.00–1.12 in boys and 0.95, 0.89–1.02 in girls for overall hospitalization).
The hospitalization rate ratio was slightly higher in the offspring of irradiated parents than in offspring of nonirradiated parents based on population comparisons (HRR= 1.1 versus 1.0) but was unrelated to estimated radiation dose to gonads (not shown). Slightly higher HRRs were seen among offspring of female compared to male cancer survivors (HRR = 1.1 versus 1.0), whereas the survivors' age at cancer diagnosis and calendar year at diagnosis did not have any measurable effect on the risk. Further, no deviation in risk was observed according to the main category of cancer in the survivor except for retinoblastomas being hereditary.
In this population-based cohort study with unbiased ascertainment of cancer cases, complete identification of their children and children of siblings, selection of population comparisons, and a nationwide registration of hospitalizations, few differences in hospitalization histories were seen among offspring in the 3 cohorts. A 6-fold excess risk of hospitalization for malignant tumors in offspring of survivors could largely be explained by rare but well known hereditary cancer syndromes, e.g., retinoblastoma, and part of the excess risk of hospitalization for benign tumors might also be explained by an underlying genetic susceptibility or by increased surveillance of children born to survivors. The cancer family clusters accord with previous findings of cancer risk in the next generation based on data from the Nordic cancer registries.16 The overall hospitalization rate ratio being slightly higher in boys than in girls born of survivors based on population comparisons and also reported in offspring of siblings might be a chance finding due to multiple comparisons.
The study has several potential limitations. A relatively crude characterization of organ doses was used in the analysis of irradiated survivors. The approximate gonadal dose categories, however, were validated against more accurate doses estimated on the actual radiotherapy schemes12 for a subset of overlapping survivors also enrolled in the parent ongoing study (www.gcct.org) with good concordance (more details of this approach can be found in Winther et al.13). Risks associated with chemotherapeutic agents could not be examined because cancer registry records were incomplete. No significant risk of hospitalizations, however, was observed among survivors of leukemia, who likely received intense chemotherapy administrations, and previous studies have found no association between chemotherapy and adverse pregnancy outcomes.17 Further, while we were unable to quantify the chemotherapy administered, it is clear that many cancer survivors received such treatments, whereas the population and the sibling comparison group did not. Thus, the interpretation that the curative treatments for cancer survivors did not adversely affect their children's morbidity patterns is generally valid although we were unable to quantify the extent of the chemotherapy received.
The approach in this study using hospitalizations to detect genetic alterations is not as optimum as clinical examinations or perhaps other approaches and in general, we were able to look only at relatively broad diagnostic categories. Nonetheless, this approach provides a comprehensive and unbiased way to address the possibility that the children of cancer survivors may be in some way less fit than the children of their siblings or those in general population. It was reassuring, however, that we were able to detect an increased rate of hospitalizations for neoplasms, which was anticipated based on the known association with cancer syndromes. Finally, not all morbidity in offspring might be related to genetic alterations. Certain outcomes included in this study such as prematurity and low-birth-weight but also other outcomes previously studied such as spontaneous abortions, stillbirths and certain malformations e.g., foot anomalies might be a result of the direct effect of pelvic radiation on uterine function.6, 13, 18
Another weakness is lack of information on potential confounders, the most important being smoking habits. The child of a parent who smokes is at approximately twice the risk of having a serious respiratory infection such as bronchitis or pneumonia in early life that requires hospitalization.19 Two large studies from the Childhood Cancer Survivor Study (CCSS) in the USA20 and the British Childhood Cancer Survivor Study (BCCSS) in the UK,21 however, reported a prevalence of smoking among survivors being substantially less than that in the general population. Also information on demographic and socioeconomic indicators such as cohabitation status and level of education which might influence on the threshold of hospitalization would have been valuable.
Trans-generational effects have traditionally been considered mutational changes in DNA sequences (gene mutations, changes in chromosome number, deletion or rearrangement of chromosomal material) induced by mutagenic exposures taking place during the production of the germ cells.22 Such permanent damage could be manifested as impaired fertility in the exposed parents, adverse pregnancy outcomes, and health problems in the offspring. Other effects such as microdeletions and induction of changes at repeat sequences in DNA (minisatellites, microsatellites and tandem repeats) are still other types of genetic damage that could impair health.
Reassurance is provided in several population-based epidemiological studies addressing spontaneous abortions13 or induced abortions due to fetal abnormality23 in female cancer survivors and congenital malformations,6 cancer,16 chromosome aberrations,5 or sex ratio alterations24, 25 in the offspring of childhood cancer survivors. Similarly, radiation didn't seemed to increase germline minisatellite mutations in children born to survivors of childhood cancer in a Danish pilot study.26
The results of this study as of previous studies of cancer survivors and their offspring suggest that the agents and doses to which these survivors have been exposed do not induce transmissible mutations in human spermatogonial stem cells or resting oocytes at a frequency high enough to be detected over the background of spontaneous mutations27 or that the mammalian organism can eliminate serious abnormalities or lethal mutations before conception or early in pregnancy and, therefore, result in surviving offspring that have a normal or background incidence of defects and disorders.28 Because of the increasing numbers of cancer survivors experiencing parenthood, however, continued monitoring of genetic risks and overall well-being among their offspring is warranted.
None of the funding sources had involvement in the conduction of this study, except for Dr. Boice, who is a Professor of Medicine, Vanderbilt University (and Principal Investigator on the National Cancer Institute grant). The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication. The authors thank data manager Ms Andrea Bautz, Institute of Cancer Epidemiology, for preparing the data set for analysis and Ms Catherine Kasper and Ms Rita Weathers, University of Texas, M.D. Anderson Cancer Center, Houston, Texas, for their contribution to dose estimates.