1 Department of Epidemiology, Netherlands Cancer Institute, Plesmanlaan 121, 1066 CX Amsterdam, 2 Department of Obstetrics and Gynecology, University Hospital Rotterdam, Postbus 2040, 3000 CA Rotterdam, 3 Department of Pediatric Oncology, Emma Kinderziekenhuis, Academic Medical Center, Meibergdreef 9, 1105 AZ Amsterdam, The Netherlands
![]() |
Abstract |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Key words: cancer incidence/children/IVF/long-term follow-up/ovulation induction
![]() |
Introduction |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Perinatal outcomes after IVF not only differ from those after spontaneous conception because of the high frequency of multiple pregnancies, but also because these pregnancies have higher complication rates. This may be related partly to the fact that mothers who achieve pregnancy after IVF are older and have lower parity. IVF pregnancies are at higher risk of pre-term deliveries and children conceived by IVF have been reported to have lower birthweights (Doyle et al., 1992; Tanbo et al., 1995
; Bergh et al., 1999
; Koudstaal et al., 2000
).
Several reports have been published on cancer in children born after IVF or fertility drug use (Steensel-Moll et al., 1985; Kobayashi et al., 1991
; Schwartzbaum, 1992
; Michalek et al., 1996
; Roman et al., 1997
; Doyle et al., 1998
; Bruinsma et al., 2000
; Lerner-Geva et al., 2000
). Unfortunately, many of these publications have been case reports, which cannot constitute proof of a causal relationship. Casecontrol studies on childhood cancer in relation to the use of fertility drugs (Steensel-Moll et al., 1985
; Schwartzbaum, 1992
; Michalek et al., 1996
; Roman et al., 1997
) deal with considerable potential for recall bias. Cohort studies following children born after IVF may overcome these problems, but those conducted have not been large enough, suffered from short follow-up and lacked an adequate control group (Doyle et al., 1998
; Bruinsma et al., 2000
; Lerner-Geva et al., 2000
).
According to the initiationpromotionprogression model derived from experimental chemical carcinogenesis, some cancers may be initiated during the early stages of fetal development. A fully recognized and established transplacental chemical carcinogen is the synthetic oestrogen diethylstilboestrol, which was widely used to prevent spontaneous abortion and was shown to develop clear cell adenocarcinoma of the vagina and cervix in female offspring (Herbst et al., 1971). Although parental exposures during pregnancy may cause both morphologic and carcinogenic changes similar to those observed in mutant mice, this has not been demonstrated in humans so far (Bolande, 1999
).
Any suspicion that fertility drugs may have long-term health effects on the offspring emphasizes the need to examine these effects more extensively. In four years, one of the authors (J.de K.) observed at least eight malignancies in children born after hormone stimulation for IVF and 11 malignancies in children born after insemination techniques and/or other use of fertility drugs. Subsequently, we examined the association between fertility techniques such as IVF and specific childhood cancers in a large nationwide cohort study of women diagnosed with subfertility problems.
![]() |
Materials and methods |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Women alive on January 1, 1997 were mailed a questionnaire to obtain information on gynaecological disorders before and after subfertility treatment, reproductive risk factors for hormone-related cancers and a number of other variables. Cohort members were traced through the Dutch Telephone Service Company and searches at municipal resident registries to assess vital status. From the initial 26 428 women, 1105 women (4.2%) were not approached because of death (n = 97), unknown, incomplete or foreign addresses (n = 495), emigration (n = 447), or specific other reasons related to privacy (n = 66). As a result 25323 women received the 23 page study questionnaire, an information letter and a study brochure. An extra form was attached to the questionnaire asking each participating subject for written informed consent for data abstraction from the medical records and linkage with the Netherlands Cancer Registry. The study information letter was signed by the attending gynaecologist or, if he/she had left, the current head of the IVF department. After 46 weeks, non-responders were sent a reminder. Non-responders to the second letter were subsequently approached by telephone (once).
For the purpose of the present research question the study population consisted of all offspring of women who returned our questionnaire (response rate 66.9%). Initially 29148 pregnancies were reported in the 16284 questionnaires that were returned. As shown in Table I, we excluded all reported miscarriages (n = 10291), all infants who were born dead (n = 524) and all children who had not yet been born at the time of the interview (n = 404). We considered the offspring eligible for analysis if the duration of gestation was at least 26 weeks. If the duration of gestation was 24 or 25 weeks, we checked for information on the vital status of the child at birth. In case the mother had not indicated whether or not the child was born alive, the child had to weigh at least 1000 g to be included in the cohort. We excluded 78 pregnancies because it was unclear whether the child was born alive, due to missing information on duration of gestation and birthweight. Of the 17851 infants who where born alive, we did not include children for whom gender (n = 298), birth date (n = 236) or exposure status (n = 299) were unknown. We furthermore excluded eight children with a 15th birthday before January 1, 1980 and 10 children who died before January 1, 1980. Thus, the total cohort for this analysis consisted of 17000 children; 9479 children were conceived by ART (IVF, inseminations, and fertility drug use not for IVF or inseminations) and 7521 children were conceived naturally.
|
For each reported child, the questionnaire completed by the study participants provided detailed information on the method of conception (IVF, insemination, use of hormones other than for IVF or insemination procedures, or natural conception), the duration of gestation in weeks, date of birth, gender, birthweight and months of breast feeding.
Information on cancer in the offspring was gathered in two separate sections of the questionnaire. Self-reported serious health problems of children were ascertained from the question: `Were there, at any time, serious health problems diagnosed in one or more of your children? If so, which health problems and in which child?' Since all the answers to this open-ended question were entered into the database exactly as they were written in the questionnaires, we were able to search for the word `cancer' and similar related terms (tumour, leukaemia, Hodgkin, neuroblastoma, Wilms', etc.). In addition, we inquired about any cancer in the first-degree relatives of the participating women in the study, and if so, which relative was affected, the type of cancer and the age at diagnosis. In case the mother reported cancer in her child as an answer to one or both of these questions, we contacted her again by letter to verify the diagnosis and to obtain permission to get more specific information from the physician treating her child. All mothers but one responded to our request for more information. The paediatrician was asked to provide information on the date of diagnosis, morphology, clinical stage and pathological stage.
Statistical analysis
The 2 test was used to compare categorical variables between the exposed and unexposed group. Year of entry into the cohort was stratified according to quartiles, based on distribution in the unexposed cohort (198019831984199019911994 and
1995). The age of the child in completed years at the end of follow-up was stratified into 2-year age groups. Method of conception was classified into the following categories: natural conception, use of fertility drugs (not for insemination or IVF treatment), inseminations, and IVF treatment. Maternal age in completed years at the infant's birth was stratified into 5-year age groups, except for the lowest (
24 years) and the highest (
40 years) age groups. Other variables included duration of the gestation [stratified into 2-week groups, except for the lowest (
32 completed weeks) and the highest group (
41 completed weeks)], birthweight [stratified into 500 g groups, except for the lowest (
1499 g) and the highest group (
4000 g)], number of siblings (no siblings, 12 siblings and
3 siblings) and multiple pregnancy (singleton, twin and
triplet). The Student t-test was used to compare the mean differences of continuous variables such as age of child at follow-up, maternal age, duration of the pregnancy, birthweight, number of siblings and duration of the follow-up.
A comparison was made between cancer incidence in our cohort of offspring of mothers who received subfertility treatment and cancer incidence in the general population. In this person-years type of analysis, the ratio of the observed (O) and expected (E) number of cancers in the study population was determined. The date of entry into the cohort was defined as the date of birth, or January 1, 1980 if the child was born before this date. Each child contributed person-years at risk from the date of entry until the earliest of the following events: 15th birthday, date of returning the questionnaire, date of cancer diagnosis or date of death. Since we did not include a specific question as to whether a child had died, we had to assume that, unless specified in the questionnaire, the child was still alive at the end of follow-up. For 116 children, the mother reported that the child had died during follow-up. Taking into account the person-years of observation in the cohort (by age, gender and calendar period), expected numbers of cancer were computed with the use of age-, sex-, and calendar period-specific cancer incidence rates from the Eindhoven Cancer Registry up to 1990 (Muir et al., 1987), and from the Netherlands Cancer Registry for the period 19901997 (van der Sanden et al., 1995
; Parkin et al., 1997
). Cancer incidence data for the whole country were not available for the total study period. The registration area of the Eindhoven Cancer Registry covers 6% of the Dutch population and is situated in the Southeastern part of the Netherlands. Reference rates of recent years from this area have been found to be comparable with those of national incidence rates. The confidence limits of O/E were obtained with the use of the Poisson distribution of O (Pearson and Hartley, 1976
). O/E ratios were calculated for all children, and separately by exposure to fertility drug treatment, gender and follow-up period. The Cox proportional hazards model was used to further quantify the effects of ART treatment, maternal age and birthweight on the risk of childhood cancer (Cox, 1972
). In this model, comparisons of childhood cancer risk within our cohort are made, adjusting for gender and follow-up time.
![]() |
Results |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
|
For the total cohort of 17000 children, 20 cases of childhood cancer were reported in the questionnaire. The average duration of follow-up was 6.0 years (4.6 years in the exposed cohort and 7.8 years in the unexposed cohort). We excluded three cases because the age at diagnosis of the cancer was 15 years. Furthermore, we excluded one case that, after verification, was not classified as malignant. Thus, the total number of cancers included in the analysis was 16. For 11 cases we were able to obtain medical verification from the paediatricians. For four cases the additional information regarding diagnosis and treatment given by the mother, after contacting her by letter, was specific enough to classify the diagnosis into the categories of cancer as listed in Table III
. For only one case, no extra information was available. Based on the information in the questionnaire (tumour of the eye nerve, of which the child died), we classified this tumour as a sympathetic nervous system tumour located in the eye. In summary, there were four cases of acute lymphoid leukaemia and one case of unspecified leukaemia, one B-cell non-Hodgkin lymphoma, one T-cell malignant lymphoma, one ependyoma, one astrocytoma, one neuroblastoma, one sympathetic nervous system tumour located in the eye, one Wilms' tumour, one clear cell sarcoma of the kidney, one osteosarcoma, one soft tissue sarcoma, and one gonadal germ cell tumour. Eleven of the 16 children died from their childhood cancer. One child with leukaemia was also born with Down's syndrome.
|
When defining the children's cohort, we assumed that all children were born to women with a history of subfertility, with or without infertility treatment. Potential misclassification could occur when children were conceived before their mothers/parents developed fertility problems. Secondary infertility is common and subfertility cannot strictly be considered as fixed in time in this context. Therefore we included, in a separate analysis, only those children who were conceived after the parents presented to the IVF clinic. Among the 8711 exposed children, six tumours were observed in 36658 person-years, against 6.1 expected (SIR = 1.0; 95% CI 0.42.1). Among naturally conceived children (n = 4214), three malignancies occurred against 3.4 expected (SIR = 0.9; 95% CI 0.22.6). The overall SIR for both groups combined was 0.9 (95% CI 0.42.1).
As shown in Table IV, seven cases of childhood cancer occurred in the exposed cohort, whereas 7.1 were expected (SIR = 0.98; 95% CI 0.42.0). The number of observed leukaemia cases in this group was three (SIR = 1.3; 95% CI 0.33.7). The total number of observed cancer cases in the unexposed cohort was nine, rendering a SIR of 1.1 (95% CI 0.52.0). Six of the seven children with cancer in the exposed group were conceived with standard IVF. One child was conceived in the seventh cycle of clomid administration. All the mothers of children conceived after IVF received human menopausal gonadotrophin and/or FSH and IU HCG. Furthermore, mothers received various protocols to support the luteal phase of the menstrual cycle. In addition, at least three mothers used gonadotrophin-releasing hormone (GnRH) agonists to down-regulate the responsiveness of the pituitary to GnRH. One child was part of a quadruplet pregnancy, and two children had twin siblings. More girls than boys were affected with childhood cancer. In the exposed cohort, five cancer cases occurred among female offspring (SIR = 1.6; 95% CI 0.53.8) and two among male offspring (SIR = 0.5; 95% CI 0.01.8). Although in absolute numbers more cancer cases occurred before the age of 5 years in the exposed cohort and the unexposed cohort, this was within the expected ranges, both for all cancer sites combined and for leukaemia (Table IV
).
|
|
![]() |
Discussion |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
So far at least four cohort studies on childhood cancer risk in children born after ART have been published. Doyle et al., compared records from the register of 2507 children conceived by ART in Britain between 1978 and 1991 with the National Registry of Childhood Tumours (Doyle et al., 1998). Only two cases of cancer were identified compared with 3.5 cancers expected (SIR = 0.6; 95% CI 0.72.1). The large Swedish study published by Bergh et al. collected data from all IVF clinics in Sweden and compared the obstetric outcomes of IVF babies (n = 5856) born between 1982 and 1995 with all babies born in the general population (Bergh et al., 1999
). Four children developed cancer, whereas 3.6 cases were expected based on general population rates. In an Australian cohort, 5249 IVF children were linked with a population-based cancer registry to determine the number of cancer cases that occurred during follow-up. A total of 4.3 cases of cancer were expected while six were observed, giving a non-significantly increased risk of cancer in the IVF group as compared with the general population (SIR = 1.4; 95% CI 0.63.1) (Bruinsma et al., 2000
). A small Israelian study reported on a cohort of 332 children born after IVF. Although 1.7 cases of childhood cancer were expected in this cohort, no cases were observed (Lerner-Geva et al., 2000
). All of these cohort studies found that ART pregnancies carried no increased risk of childhood cancer in comparison with the general population. However, the results were based on very small numbers and short follow-up. Furthermore, none of the published studies was able to make a direct comparison between ART children and an appropriate control group.
A few casecontrol studies have suggested an increased risk of neuroblastoma (Kramer et al., 1987; Michalek et al., 1996
) and leukaemia (Steensel-Moll et al., 1985
; Roman et al., 1997
) in the offspring of women treated with fertility drugs. The epidemiology of neuroblastoma suggests that prenatal exposures may be important aetiologic factors in this disease. Kramer et al. reported significantly elevated odds ratios (ORs) of neuroblastoma in association with maternal use of sex hormones three months prior to or during pregnancy (OR = 2.25) (Kramer et al., 1987
). Another study by Michalek et al. among 183 histologically confirmed neuroblastoma cases aged 04 years and matched controls reported a significant increase in risk among the offspring of women who used hormones related to infertility (OR = 10.4, 95% CI 1.289.9) (Michalek et al., 1996
). Schwartzbaum reported on 1270 children diagnosed with cancer between 1979 and 1986, including 104 children with neuroblastoma (Schwartzbaum, 1992
). A total of eight mothers reported using fertility agents from one year before the pregnancy to the time of birth, revealing no increased risk of neuroblastoma. The results of large cohort studies among IVF children, including our findings, do not show an increased risk for neuroblastoma compared with the general population rates so far (Bergh et al., 1999
; Bruinsma et al., 2000
). In fact, the one child with neuroblastoma observed in our cohort was conceived naturally.
The Dutch casecontrol study by Steensel-Moll et al. included 519 children with acute lymphocytic leukaemia and matched 507 controls (Steensel-Moll et al., 1985). A history of fertility problems (OR = 6.0; 95% CI 0.938.2) was more frequently reported among mothers of cases. The related use of `drugs to maintain pregnancy' was 1.9 (95% CI 1.03.5). A medical record-based study conducted in England included 177 leukaemia and non-Hodgkin's lymphoma patients who were diagnosed before the age of 30 years (92% of cases were diagnosed at the age of
24 years) and 354 matched controls (Roman et al., 1997
). The RR associated with hormonal treatment for infertility of the mother was 2.7 (95% CI 0.611.9). The RR associated with reported infertility investigations at the time of the first antenatal clinic visit for the index pregnancy was 2.1 (95% CI 0.94.6) (Roman et al., 1997
). The study by Schwartzbaum reported no increased risk of fertility agents in the 629 children diagnosed with leukaemia (Schwartzbaum, 1992
). In our cohort, we found no increased risk for leukaemia in children born to women treated for subfertility. However, this result was based on only five cases of leukaemia, two of which were in the unexposed group.
When interpreting our results, the strengths and limitations of the study design need to be considered. The strengths of our study include the relatively large study size and the possibility of medical verification of exposure and reported outcomes. A disadvantage of our study is that cancer incidence was based on self-reported information that was subsequently medically verified. When assessing the incidence of cancer in a cohort, linkage with a national cancer registry is preferable compared with self-reported information. Unfortunately, since the OMEGA-cohort was not designed to study specific disorders in children, we did not obtain written permission from the parents to compare our cohort of children with the Netherlands Cancer Registry. However, in the questionnaire, two different questions were included in which mothers were asked to report serious health problems affecting their child(ren). Because of the seriousness of the disease under investigation, we do not expect under-ascertainment of childhood cancer among participating women.
Because information on childhood cancer in our study was collected retrospectively, selection and/or reporting bias may have occurred. In two hospitals we collected data for all eligible women in the cohort (instead of only for responding women). In these two hospitals, 73% of the women with children (as mentioned in the medical files) responded to the questionnaire, whereas only 60% of the women without children responded. Although the overall response rate of the initial OMEGA-cohort is 66.7%, the actual response rates for women with children (the relevant subgroup since their offspring was included in the OMEGA's children cohort) is 6% higher. If the mothers of children with cancer in the exposed group were more inclined to respond to the questionnaire, or to report the occurrence of childhood cancer, this might have led to an overrepresentation of childhood cancer in the exposed group. However, in both the exposed and unexposed group, the risk of childhood cancer was similar to the risk seen in the general population, rendering selection bias or differential reporting by exposure unlikely.
To date, ~6070% of all children affected with childhood cancer are being cured (Robison, 1993). In our cohort, 11 of the 16 children died of their disease. The number of deaths is higher than expected compared with general population rates. We can think of three possible explanations for this result. First, it may be a change finding. Second, some women with a living child with cancer may have been unwilling to participate in the study because they were, at the time of the questionnaire, in the process of taking care of their sick children. It appears unlikely that this would have differentially affected the response rates of women with children conceived by ART and those with naturally conceived children. Another, less likely, explanation would be that more aggressive tumours may have occurred in our cohort of children of subfertile women.
In conclusion, in this study we found no evidence that ART, or more specifically IVF treatment, is associated with an increased risk of cancer in children. Although our study included a large number of children, cancer during childhood is rare and our findings are based on small numbers of cases. The statistical power of our study was sufficient to detect a 2.3 fold increase in the risk of childhood cancer in the exposed group with 80% probability (Breslow et al., 1987). Lower elevations of risk could not be detected with the numbers available. Furthermore, because almost 75% of the children in the exposed group were
5 years, longer follow-up of children conceived by ART is necessary to exclude long-term adverse effects of ART treatment. Definite conclusions regarding the risk of childhood cancer from ART can only be drawn when the observed numbers allow risk estimation for different tumour sites and control for various confounders. Different fertility treatments that include ovulation induction may have different potential effects on gametes and embryos. In order to increase the number of childhood cancer cases needed for such analyses, much longer follow-up time is required, and pooling of cohort studies from different countries is recommendable.
![]() |
Acknowledgements |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
![]() |
Notes |
---|
4 To whom correspondence should be addressed. E-mail: fvleeuw{at}nki.nl
![]() |
References |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Bolande, R.P. (1999) Prenatal exposures and childhood cancer. Pediatr. Dev. Pathol., 2, 208214.[ISI][Medline]
Breslow, N.E. and Day, N.E. (1987) Statistical methods in cancer researchvol 2: the design and analysis of cohort studies. IARC, Lyon.
Bruinsma, F., Venn, A., Lancaster, P. et al. (2000) Incidence of cancer in children born after in-vitro fertilization. Hum. Reprod., 15, 604607.
Cox, D.R. (1972) Regression models and life tables. J. Roy. Statist. Soc. Ser. B., 334, 187202.
de Jong-van den Berg, L.T.W., Cornel, M.C., van den Berg, P.B. et al. (1992) Ovulation-inducing drugs: a drug utilization and risk study in the Dutch population. Int. J. Risk Safety Med., 3, 99112.
Doyle, P., Beral, V. and Maconochie, N. (1992) Preterm delivery, low birthweight and small-for-gestational-age in liveborn singleton babies resulting from in-vitro fertilization. Hum. Reprod., 7, 425428.[Abstract]
Doyle, P., Bunch, K.J., Beral, V. et al. (1998) Cancer incidence in children conceived with assisted reproduction technology. Lancet, 352, 452453.[ISI][Medline]
Herbst, A.L., Ulfelder, H., and Poskanzer, D.C. (1971) Adenocarcinoma of the vagina. Association of maternal stilbestrol therapy with tumor appearance in young women. N. Engl. J. Med., 284, 878881.[ISI][Medline]
Kobayashi, N., Matsui, I., Tanimura, M. et al. (1991) Childhood neuroectodermal tumours and malignant lymphoma after maternal ovulation induction. Lancet, 338, 955955.
Koudstaal, J., Braat, D.D., Bruinse, H.W. et al. (2000) Obstetric outcome of singleton pregnancies after IVF: a matched control study in four Dutch university hospitals. Hum. Reprod., 15, 18191825.
Kramer, S., Ward, E., Meadows, A.T. et al. (1987) Medical and drug risk factors associated with neuroblastoma: a casecontrol study. J. Natl Cancer Inst., 78, 797804.[ISI][Medline]
Lerner-Geva, L., Toren, A., Chetrit, A. et al. (2000) The risk for cancer among children of women who underwent in-vitro fertilization. Cancer, 88, 28452847.[ISI][Medline]
Michalek, A.M., Buck, G.M., Nasca, P.C. et al. (1996) Gravid health status, medication use, and risk of neuroblastoma. Am. J. Epidemiol., 143, 9961001.[Abstract]
Muir, C., Waterhouse, J.A. and Mack, T. (eds) (1987) Cancer Incidence in Five Continents Vol V. IARC Scientific Publication No. 88, Lyon, France.
Parkin, D.M., Whelan, S.L., Ferlay, J. et al (eds) (1997) Cancer Incidence in Five Continents Vol VII. IARC Scientific Publication No. 143, Lyon, France.
Pearson, E.S. and Hartley, H.O. (eds) (1976) Biometrika Tables for Statisticians. Biometrika Trust, London, England.
Robison, L. (1993) General principles of the epidemiology of childhood cancer. In Pizzo, P.A. and Poplack, D.G. (eds) Principles and Practice of Pediatric Oncology. J.B.Lippincott, Philadelphia, pp. 310.
Roman, E., Ansell, P. and Bull, D. (1997) Leukaemia and non-Hodgkin's lymphoma in children and young adults: are prenatal and neonatal factors important determinants of disease? Br. J. Cancer, 76, 406415.[ISI][Medline]
Schwartzbaum, J.A. (1992) Influence of the mother's prenatal drug consumption on risk of neuroblastoma in the child. Am. J. Epidemiol., 135, 13581367.[Abstract]
Steensel-Moll, H.A., Valkenburg, H.A., Vandenbroucke, J.P. et al. (1985) Are maternal fertility problems related to childhood leukaemia? Int. J. Epidemiol., 14, 555559.[Abstract]
Tanbo, T., Dale, P.O., Lunde, O. et al. (1995) Obstetric outcome in singleton pregnancies after assisted reproduction. Obstet. Gynecol., 86, 188192.
van der Sanden, G.A., Coebergh, J.W., Schouten, L.J. et al. (1995) Cancer incidence in The Netherlands in 1989 and 1990: first results of the nationwide Netherlands cancer registry. Coordinating Committee for Regional Cancer Registries. Eur. J. Cancer, 31A, 18221829.
Submitted on April 17, 2001; accepted on August 7, 2001.