Reproductive Outcomes in Male Childhood Cancer Survivors:  A Linked Cancer-Birth Registry Analysis FREE

Chemotherapy, radiotherapy, and surgical treatment for cancer may impair future reproductive potential, and concerns about fertility and the health of any progeny are increased among survivors of childhood cancers compared with siblings.^1- 2 Most prior reports are based on institutional case series that include self-reported reproductive outcomes. There have been fewer studies using population-based data, and most were only able to examine a limited number of outcomes: fertility rate, sex ratio, and rates of malformations and cancer among progeny.^3- 8 In general, these studies suggest that, although male survivors are less likely than expected to father children, the pregnancies they father tend to have few other complications, and adverse outcomes among progeny are not increased.

We identified a population-based sample of male childhood cancer survivors from 4 US regions and linked their cancer registry records to state birth certificate registries to determine the proportion who subsequently fathered live births. We then compared the occurrences of selected pregnancy conditions and infant outcomes between partners of male cancer survivors and those of a population-based comparison group identified from birth records.

Human subject protection committee approval was received by the appropriate institutions and state departments of health before this study began. Methods used to identify participants and link data are described in detail in an accompanying article.⁹ Briefly, incident cancer cases occurring among males younger than 20 years, newly diagnosed with cancer (malignant and in situ), were identified from 4 population-based cancer registries participating in the National Cancer Institute's Surveillance, Epidemiology and End Results (SEER) Program during the following time intervals: the Cancer Surveillance System of Western Washington in Seattle, 1974-1995; the Karmanos Cancer Institute of Wayne State University in Detroit, Michigan, 1973-2000; the Utah Cancer Registry at the University of Utah, 1973-1998; and the SEER registry in Atlanta, Georgia, 1975-2000. Aside from the Utah Cancer Registry, these registries are not statewide but only include the named metropolitan region plus surrounding counties (registry details may be found at http://seer.cancer.gov/registries/index.html). The registries provided data on patients' demographic and tumor characteristics as well as the initial course of treatment (any chemotherapy, any surgical treatment, any radiotherapy, and nonoverlapping combinations). Childhood cancer diagnoses were categorized using the International Classification of Childhood Cancer,¹⁰with categories corresponding to neuroblastoma and related tumors, embryonal renal and hepatic tumors, and retinoblastoma collapsed into a single embryonal tumor category because of small sample sizes.¹¹ The anatomical primary cancer site also was categorized as to whether it occurred within the pelvis. Cancer relapse information was unavailable.

Birth certificate data from all 4 states were linked to each cancer patient's registry record to identify the live-born delivery occurring in closest temporal proximity following the participant's cancer diagnosis date for these available years: Washington, 1974-2001; Utah, 1973-2001; Michigan, 1975-2001; and Georgia, 1980-2000. Routine linkage strategies varied by state, with available linkage variables including patients' first and last names, sex, birth dates, birth place (Utah only), race/ethnicity (Georgia only), and social security number (Utah, Michigan, and Georgia). A total of 483 potential subjects were identified from the 4 regions and were linked to the birth records in each state. Records of live-born deliveries that occurred before the cancer diagnosis were not used in this study.

For comparison, men who fathered infants born during the same year were randomly selected from among the remaining birth records at a comparison to case subject ratio of 4:1 in Michigan and 10:1 in the other 3 states. These also were frequency matched on the cancer survivor's age at delivery (5-year intervals from <20 to ≥40 years) and race/ethnicity. On examination of the linked file, it was determined that some potential cases were ineligible and were subsequently excluded: 12 had benign lesions and 1 had a basal cell skin tumor. In addition, 2 records associated with comparison subjects were of fetal deaths and were excluded because the analysis focused on live births only. This resulted in 470 cancer survivors and 4150 comparison subjects being included in the final analysis.

Female partner outcomes that could be evaluated using birth records included delivery type, maternal anemia, diabetes, and preeclampsia. Infant outcomes included birth weight (<2500, 2500-3999, or ≥4000 g), gestational age (<37, 37-41, or ≥42 weeks), small-for-gestational-age status (defined as <10% birth weight for gestational age and sex based on a representative national sample¹²), the presence of any malformation, 5-minute Apgar score of less than 7 (unavailable in Michigan), and infant death at younger than 12 months (unavailable in Georgia). Other information available for most records included the female partner's prenatal smoking status, marital status, number of prior pregnancies and births, and when prenatal care was initiated. No information on assisted reproductive techniques was available.

The number of cancer survivors in each region who were linked with birth certificates and the total number of cases ascertained in each SEER region during the same time period (SEER*Stat database, version 6.1.4, released in 2005; National Cancer Institute, Bethesda, Maryland) were used to calculate the proportion of survivors in each region identified with subsequent live births. The distribution of selected parental characteristics and maternal and infant outcomes was described for cancer survivors and comparison subjects. Relative risks (RRs) estimated using stratified Mantel-Haenszel methods were used, with results similar to those produced by logistic regression, log-binomial, or Poisson models.¹³ All RRs were adjusted for state, the frequency-matched variables, and maternal age and parity. Other variables considered for their possible role in the relationships of interest included: infant sex and maternal race/ethnicity, prenatal smoking status, marital status, and duration of prenatal care. However, except where noted, adjustment for these variables did not meaningfully alter the RR estimates, and only those variables for which inclusion resulted in such change were retained in the analyses. Sensitivity analyses in which births occurring within 9 months of cancer diagnosis, multiple-gestation births, and multiparous partners were excluded showed similar results. Analyses were conducted using Stata statistical software, version 9 (StataCorp, College Station, Texas).

General diagnostic and treatment characteristics of male cancer survivors (Table 1) were similar across SEER regions, with several exceptions: a greater proportion of patients from Detroit had 11 to 30 years elapse between diagnosis and subsequent delivery (69.9% vs 34.1%-38.5% in other regions); the prevalence of skin cancer ranged from 2.9% in Detroit to 12.3% in Utah; and the prevalence of selected cancer treatments varied (chemotherapy ranged from 7.8% in Utah to 18.9% in Seattle; surgical treatment plus radiotherapy ranged from 4.9% in Detroit to 17.6% in Atlanta; and any surgical treatment ranged from 51.5% in Detroit to 70.1% in Utah). Overall, the proportion of childhood cancer patients identified from birth records as having fathered a live birth ranged from 3.7% in Detroit to 11.3% in Utah, with an overall median time from cancer diagnosis to fatherhood of 10 (range, 0-28) years.

Compared with all childhood cancer cases ascertained by the SEER Program in the 4 regions during the study period (per the SEER*Stat database), the subset of patients included in this study was more likely to be diagnosed in an earlier era (before 1990, 84.9% vs 59.6%) and at an older age (≥10 years, 86.4% vs 50.3%) and was less likely to have those cancers associated with younger age at diagnosis (leukemia, 10.9% vs 24.9%; embryonal tumors, 5.3% vs 13.0%).

Paternal race/ethnicity, age, and year of delivery, which were frequency matched by design, were similar for cancer survivors and comparison subjects (Table 2). The median age of cancer survivors and comparison subjects at delivery was 25 (range, 16-40) and 24 (range, 16-47) years, respectively. Both groups also were similar with respect to female partner's age, race/ethnicity, prenatal smoking status, and the proportion with multiple-gestation births (2.6% vs 2.1%). However, greater proportions of cancer survivors had female partners who were recorded as being unmarried, primigravida, and nulliparous.

Overall, female partners of cancer survivors had similar risks of selected pregnancy-related conditions compared with partners of comparison subjects (Table 3). Among infant outcomes, the male to female offspring ratio did not differ significantly between survivors and comparison subjects (1.09 vs 1.04, respectively; RR for having male offspring, 1.03 [95% confidence interval (CI), 0.93-1.14]). Cancer survivors had a borderline risk of fathering infants with a birth weight of less than 2500 g (RR, 1.43 [95% CI, 0.99-2.05]) but no risk of having infants weighing less than 1500 g (1.16 [0.46-2.93]). These risk estimates were not affected by adjustment for gestational age and maternal factors such as prenatal smoking status, preeclampsia, or diabetes. Offspring of cancer survivors did not have an increased risk of being born at less than 37 weeks' gestation. Furthermore, although the proportion of preterm infants weighing less than 2500 g at birth was slightly greater among cases (54.8%) than comparison infants (47.7%), offspring of survivors were not at increased risk of meeting small-for-gestational-age criteria. The RRs for infant malformations and 5-minute Apgar scores of less than 7 also were not increased, and there were no infant deaths among progeny of cancer survivors.

When maternal and infant outcomes were stratified by cancer subtype and pelvic cancer location, no consistent associations were seen for cesarean section, maternal diabetes or anemia, preterm delivery, and infant malformations. However, partners of men with childhood central nervous system tumors (RR, 3.36 [95% CI, 1.63-6.90]) and leukemia (2.41 [95% CI, 1.11-5.22]) had an increased risk of preeclampsia (Table 4), even after additional adjustment for maternal diabetes. However, no other consistent associations for preeclampsia were observed among other subgroups.

Significant increased risks of fathering a low-birth-weight infant were associated with embryonal tumors (RR, 3.93 [95% CI, 1.68-9.20]) (Table 4). No secular trends were observed, except for an increased risk of low birth weight associated with earlier treatment era (pre-1980) that diminished over subsequent decades. Young age at diagnosis (<5 years) and greatest elapsed time since diagnosis (>10 years) also were associated with birth weight of less than 2500 g. Treatments involving any chemotherapy (RR, 1.96 [95% CI, 1.22-3.17]) or any radiotherapy (1.95 [95% CI, 1.14-3.35]) were associated with a greater risk of low birth weight, but exposure to surgical treatment alone was not. Patients who had primary tumors in the abdomen treated with any radiotherapy also were at risk (RR, 3.38 [95% CI, 1.49-7.68]). Risk estimates for combination therapies were more variable but, in general, those containing chemotherapy tended to be greater than those without chemotherapy, with the greatest risk associated with exposure to all 3 modalities (RR, 3.47 [95% CI, 1.36-8.85]). Borderline associations for having a small-for-gestational-age infant also were seen following treatment with any radiotherapy (RR, 1.58 [95% CI, 1.03-2.42]) and combination chemotherapy with radiotherapy (2.29 [1.13-4.63]) but not with other subgroups.

Fathers with pelvic primary tumors had a male to female offspring ratio of 1.18, but the likelihood of having male offspring was not increased significantly (RR, 1.02 [95% CI, 0.84-1.26]). Infant male to female ratios were greatest for those aged 10 to 14 years at diagnosis (1.18) but were similar for all other diagnostic age categories (range, 1.06-1.07). Infant male to female ratios following any radiotherapy, any chemotherapy, and surgical treatment only were 1.23, 1.16, and 0.93, respectively, but none of these were significantly different from the ratios of comparison subjects, even after multivariable adjustment or restriction to those with pelvic tumors.

Few studies^14- 15 have examined associations between paternal cancer history and subsequent pregnancy complications among female partners and outcomes among offspring outside of malformations or cancer. In this study of a relatively contemporary cohort of male childhood cancer survivors, partners and progeny of cancer survivors were not at increased risk of most complications examined. However, we did observe an increased risk of low birth weight and preeclampsia associated with some treatment characteristics.

Based on self-reported outcomes of more than 1500 live births, the North American Childhood Cancer Survivor Study reported a 3-fold increased risk of low-birth-weight progeny among male childhood cancer survivors treated with nonalkylating chemotherapy compared with siblings.¹⁴ Risk estimates associated with pelvic radiation and alkylating chemotherapy (RR, 1.5-1.6) were not significantly increased but were similar in magnitude to our results. Gestational age was not examined in this study. A registry-based study examining several hundred Norwegian male cancer survivors diagnosed between ages 15 and 35 years did not report an increased risk of low-birth-weight progeny.¹⁵ However, given that our estimates were greatest for males diagnosed at younger than 15 years, differences in the study populations and their exposures may account for this discrepancy. Nevertheless, in neither study was there a consistent risk of progeny being delivered preterm.

Both maternal and paternal contributions to birth weight have been reported in the general population. Parents who were themselves low-birth-weight infants tend to give birth to low-birth-weight infants, independent of environmental factors.^16- 17 Although we did not know parental birth weights, there is no reason to suspect that male cancer survivors or their partners would more likely have been low-birth-weight infants themselves. Aside from hepatoblastoma (there were no cases in this study), low birth weight has not been associated consistently with an increased risk of childhood cancers.¹⁸ Additional maternal demographic and environmental factors associated with low birth weight include nulliparity, very young or older maternal age, having a prior low-birth-weight infant, lower socioeconomic status, and substance use, including tobacco exposure.¹⁹ We attempted to adjust for many of these factors, but it is possible residual confounding exists as childhood cancer survivors, particularly patients with central nervous system tumors, are more likely to be unmarried²⁰ and unemployed,²¹ although less likely to smoke.²² Nevertheless, it is interesting that risk was increased after exposure to chemotherapy and radiotherapy, but not to surgical treatment alone. Compared with maternal associations, any paternal influence on infant birth weight is more likely to be genetic rather than environmental.^17,23 There is evidence that the imprinting of fetal genes can affect fetal growth and adult health,²⁴ although there is no evidence that prior cancer therapy in the father affects imprinting of germ-line cells, particularly because epigenetic therapies would not have been widely used during the study period.

It is unclear why preeclampsia was associated with central nervous system tumors and, to a lesser degree, leukemia. Although we examined in our analysis maternal factors that can be associated with preeclampsia, such as age, nulliparity, and diabetes, it is possible these findings could still reflect residual confounding or be owing to chance. Prior studies generally have not examined preeclampsia among partners of male cancer survivors. One study of men who had undergone hematopoietic cell transplantation found that 6% of partners (n = 4) experienced preeclampsia,²⁵ similar to the prevalence in our study but also within estimated population rates.²⁶ A paternal contribution to preeclampsia has been reported in the general population, suggesting that paternally derived fetal genes are involved in pathogenesis.^27- 28 The mechanism by which paternal genes affect preeclampsia is unclear, although a role for paternally imprinted alleles also has been hypothesized.²⁷ However, as with low birth weight, it is unclear if cancer therapy affects germ-line imprinting. Although loss of imprinting is an increasingly recognized phenomena among pediatric and adult cancers, these changes typically are restricted to tumor cells, with the exception of certain rare cancer predisposition syndromes.²⁹

Previous studies have investigated possible mutagenic effects of cancer therapy on germ cells as manifested by an altered male to female progeny ratio, malformations, and miscarriages or stillbirths. In our study, the male to female progeny ratio was slightly greater among survivors (1.09) compared with comparison subjects (1.04) and was greatest among those with pelvic tumors and those exposed to any chemotherapy or radiotherapy (range, 1.16-1.23). In comparison, during the past 50 years, the US ratio has been around 1.05.³⁰ Although our estimates did not reach statistical significance, this pattern supports the hypothesis that mutagenic therapies could result in an increased male to female progeny ratio among treated fathers because of dominant lethal X-chromosome mutations. However, male to female ratios have not been increased in other studies of male childhood cancer survivors^6,8,31 and even significantly decreased in one.¹⁴

Most studies, including ours, have not reported increased risks of malformations,³² although our use of birth registry data may result in underascertainment of more subtle defects not diagnosed at birth. One birth registry study did report a 50% increased risk of malformations among progeny of adolescent and young adult male cancer survivors compared with the general population.¹⁵ An increased risk of miscarriage or stillbirths among partners of male cancer survivors¹⁴ also suggests the possibility that mutagenic exposures may affect viability of future offspring.

Our study has several limitations. Although SEER audits have shown that case ascertainment exceeds 95% and that tumor characteristics³³ and broad treatment categories (eg, chemotherapy, radiotherapy, and surgical treatment) are accurately recorded,³⁴ our data were limited to initial cancer treatment. Information about treatment for relapse was not available, and, therefore, our estimates for treatment categories contain some misclassification. The effects of this are difficult to predict because chemotherapy, radiotherapy, and surgical procedures are used for salvage treatments, but patients likely received more multimodal therapy than shown.

Although more than 99% of births in the United States are captured in birth records,³⁵ these records have some limitations. Generally, paternal characteristics are not as thoroughly recorded as maternal characteristics, particularly if parents are unmarried.³⁵ Because male cancer survivors were less likely to be currently married than comparison subjects, this may, in part, explain why the percentage of cancer survivors identified as fathers (6.7%) was lower than in other studies, although birth registries do attempt to record paternal information even if the couple is unmarried. Migration to other states would also decrease our cancer-birth registry linkages. However, at least for recent years, US Census surveys report less than 3% of people who move have moved out of state, and concern regarding health is rarely cited as the primary reason for moving.³⁶ Furthermore, migration would have affected our outcomes only if cancer survivors who moved out of state differed from those who remained, information we could not access. Finally, as the median age of survivors in this cohort was only 25 years, many survivors also are just entering reproductive age.

Nevertheless, for successful linkages, birth record characteristics such as gravidity, parity, delivery method, infant sex, birth weight, and gestational age are recorded accurately with sensitivity and specificity typically higher than 95% when compared with medical records.³⁷ However, although the specificity of maternal comorbidities such as diabetes, preeclampsia, anemia, and tobacco exposure is typically high, sensitivity can be much more variable.^37- 38 Overall, there is no reason to suspect that partners of male cancer survivors would have birth record data recorded differently than others; birth records should not be susceptible to response or recall biases.

Misattributed paternity may also be present within our population. Although one may hypothesize that use of donor sperm may be more prevalent among cancer survivors, there is little information about its prevalence, and we did not have any data on use of assisted reproductive techniques. Furthermore, the direction of any bias arising from misattributed paternity is difficult to predict. Among the general population, nonpaternity rates have varied greatly across populations and have been associated with different demographic factors in various studies.³⁹

Last, differences may exist between this cohort and the overall population of male childhood cancer cases. Despite follow-up of as long as 28 years, as in many other studies of childhood illnesses that attempt to examine outcomes in adulthood, our sample included more individuals who were diagnosed in earlier time periods and who were older at diagnosis. Cancers with increased reproductive morbidity and mortality also would be less represented in any survivor cohort. Nevertheless, for male survivors identified as fathers in state birth records, the vast majority of associated pregnancies resulting in live births were not at significantly greater risk of complications vs those of comparison subjects. However, our finding of increased low birth weight and preeclampsia associated with some diagnostic groups raises the possibility that prior cancer therapy may affect male germ cells with effects on female partners and progeny of male survivors.

Correspondence: Eric J. Chow, MD, MPH, Fred Hutchinson Cancer Research Center, PO Box 19024, Mailstop M4-C308, Seattle, WA 98109-1024 (ericchow@u.washington.edu).

Accepted for Publication: January 29, 2009.

Author Contributions: Drs Chow, Daling, Severson, and Mueller had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Daling and Mueller. Acquisition of data: Daling, Fraser, Wiggins, Mineau, Hamre, Severson, Drews-Botsch, and Mueller. Analysis and interpretation of data: Chow, Kamineni, Daling, and Mueller. Drafting of the manuscript: Chow, Daling, and Mueller. Critical revision of the manuscript for important intellectual content: Chow, Kamineni, Daling, Fraser, Wiggins, Mineau, Hamre, Severson, Drews-Botsch, and Mueller. Statistical analysis: Chow, Daling, and Mueller. Obtained funding: Daling, Wiggins, and Mueller. Administrative, technical, and material support: Fraser, Wiggins, Mineau, and Severson. Study supervision: Wiggins and Mueller.

Financial Disclosure: None reported.

Funding/Support: This study was supported by contract N01-PC-05016-20 from the National Cancer Institute.

Additional Contributions: William O’Brien provided data management and programming assistance. Cancer registry data were provided by the Cancer Surveillance System of the Fred Hutchinson Cancer Research Center (contract N01-CN-05230); Metropolitan Detroit Cancer Surveillance System of Wayne State University/Karmanos Cancer Institute (contract N01-CN-65064); Utah Cancer Registry (contracts N01-PC-35141 and N01-CN-67000); and the Metropolitan Atlanta SEER Registry of Emory University (contract N01-PC-67006). Vital statistics data were provided by the Washington State Department of Health, Center for Health Statistics; the Utah Department of Health with database support from the Huntsman Cancer Institute; the Vital and Health Record Section, Department of Community Health, Community Public Health Agency of the State of Michigan; and the Georgia Department of Human Resources, Division of Public Health, Office of Vital Records.