Reproductive history, infertility treatment, and the risk of acute leukemia in children with down syndrome
A Report From the Children's Oncology Group
Children with Down syndrome (DS) have from 10 to 20 times the risk of developing acute leukemia than the general pediatric population. There is mixed evidence for associations between reproductive history or infertility and acute leukemia among children without DS.
The authors conducted a case-control study of acute leukemia among children with DS to investigate possible risk factors in this population. From 1997 to 2002, 158 children aged <20 years with DS who had a diagnosis of acute leukemia (97 children with acute lymphoblastic leukemia [ALL] and 61 children with acute myeloid leukemia [AML]) were enrolled at Children's Oncology Group (COG) institutions. Controls with DS (n = 173) were selected from the cases' primary care clinic and frequency matched to cases on age. Telephone interviews were conducted with mothers of cases and controls assessing reproductive history, infertility, and infertility treatment.
Null results were observed overall and by subtype for reproductive factors, including previous pregnancy outcomes and contraceptive use, and for most infertility outcomes. There was an increased risk of AML among children with DS whose parents had ever tried for ≥1 year to become pregnant (odds ratio [OR], 2.22; 95% confidence interval [95% CI], 1.14–4.33). A 1-year increase in maternal age also was associated with AML (OR, 1.06; 95% CI, 1.01–1.12).
Although the questionnaire was limited in this area, the results suggested that the risk for AML may be raised in children with DS because of infertility. In that the risk of infertility, along with having a child with DS, increase with age, these results warrant more research. Cancer 2007. © 2007 American Cancer Society.
Leukemia is the most common type of childhood cancer in the United States, with an incidence rate of 4.1 per 100,000 children ages 0 to 19 years in 2002.1 Down syndrome (DS) is the leading chromosomal defect in the United States and has a national estimated prevalence of 13.65 per 10,000 live births.2 One of the few known risk factors for childhood leukemia is DS, which confers a 10- to 20-fold increased risk over baseline.3 For morphologic subtype M7 of acute myeloid leukemia (AML), it is estimated that the increased risk is 500-fold in children with DS.4 The reason for the increased risk of leukemia in children with DS is obscure. Current hypotheses include increased chromosomal instability, increased gene dosage effects because of trisomy, and translocations involving chromosome 21.5–9 Apart from inherent genetic susceptibility, children with DS also may have a greater susceptibility to environmental factors, including those that occur in utero.
Reproductive history often has been studied as a potential risk factor for childhood leukemia, with mixed results, although studies have not focused specifically on children with DS. Previous studies in children without DS have indicated no or little elevation in risk for advanced maternal age,10–19 prior fetal loss,12, 14–17, 20–25 prior contraceptive use,15, 26–28 time between births,16, 22 and birth order.12, 14–16, 20, 21, 29–31 However, those factors may have a greater impact on the development of leukemia in children with DS, because trisomy 21 may represent a first genetic “hit,” and stronger associations may be expected than those reported in studies of children without DS.
Infertility and its treatment are less studied aspects of reproductive history. A few studies have indicated that there may be an increased risk of childhood acute leukemia after use of infertility treatments including, ovulation-stimulating drugs, in vitro fertilization (IVF), and intracytoplasmic sperm injection, although this risk generally has been small and could be because of the underlying cause of infertility rather the treatment itself.32, 33 Prevalence of conception after treatment for infertility is likely to be higher among children with DS, because older maternal age is both an indication for infertility and a risk factor for DS among offspring.34, 35 Studies also have suggested that assisted reproduction could be a risk factor for chromosomal abnormalities like DS.36, 37 To our knowledge, this analysis is the first to explore reproductive factors in children with DS and the risk of leukemia.
MATERIALS AND METHODS
A detailed description of the study has been presented elsewhere.38, 39 Briefly, registration files of the Children's Oncology Group (COG) were used to identify children with DS who had a diagnosis of either acute lymphoblastic leukemia (ALL) or AML between January 1997 and October 2002 (cases). Cases were required to be aged ≤19 years at diagnosis, to have a telephone in their residence, and to have a biologic mother available who spoke English, and they must have resided in the United State or Canada at diagnosis. Permission to contact the parents was obtained from the child's primary care physician, and written consent was obtained from the child's mother. The diagnosis of DS was confirmed by central review using cytogenetic data and obstetric medical charts both for the case group and for a control group. The diagnosis of acute leukemia also was confirmed by central review using clinical data, pathologic specimens and reports, and cytogenetic data.
In total, 210 cases were identified in 116 COG institutions. Of these cases, 158 mothers (75%) completed interviews (97 mothers who had DS children with ALL and 61 mothers who had DS children with AML). Reasons for not completing the interview included maternal refusal (17%), physician refusal (5%), and not being able to locate the mother (3%).
Names and addresses of the primary care physician seen by cases before their diagnoses of leukemia were obtained after the telephone interview from the mothers of the cases. Each primary care physician was contacted and asked to provide a roster that included the age and sex of his or her patients with DS who had no history of leukemia. Then, controls were selected randomly from the rosters and frequency matched to cases by age. Both cases and controls were required to have a telephone in their home, an English-speaking mother who could be interviewed, and reside in the United States or Canada. Of the 151 clinics that were contacted, 77 provided rosters of patients with DS. Through these rosters, 726 potential controls were identified, and 329 children were selected randomly based on date of birth. Names and addresses were provided by the clinic for 215 of the random sample of controls. Of these, 173 mothers (80.5%) completed telephone interviews.
Telephone interviews were conducted with mothers of cases and controls to obtain data on medical and reproductive history, personal habits, and demographics. For this analysis, we examined reproductive history, contraceptive use, and infertility treatment. The index pregnancy and the index child refer to the case or control child. Reproductive history variables of interest included maternal age (continuous), maternal age (<30 years, 30–34 years, ≥35 years), prior fetal loss (0, 1, >2), prior molar pregnancy (yes/no), prior ectopic pregnancy (yes/no), time since last live birth to birth of the index child (none or ≤2 years, >2 years to ≤4 years, >4 years), and birth order (first, second, third or higher). Contraceptive use was defined by prior use of oral contraceptives (yes/no), prior use of implant or injectable contraceptives (yes/no), and use of the following contraception methods in the 2-year period before the birth of the index child: barrier method (diaphragm, condom, sponge, or something similar; yes/no), spermicides (jelly, cream, foam, or suppository contraceptives; yes/no), intrauterine devices (yes/no), and other birth control (yes/no). Variables of interest that may indicate infertility problems or treatment included the time required to conceive the index child (not trying, trying for ≤12 months, trying for >12 months), ever try for ≥1 straight year to become pregnant and not become pregnant (yes/no), and ever use medication (tablets or injections), surgical treatment, or other treatment to conceive (yes/no). A priori potential confounders included maternal age (continuous), mother's education (high school or less, more than high school), mother's race (white, nonwhite), and index child's sex (boy or girl).
Odds ratios (OR) and 95% confidence intervals (95% CI) were calculated from unconditional logistic regression models to examine the correlation between reproductive history and infertility exposures and leukemia. Separate models were constructed for each exposure of interest, because small sample sizes had the potential to make parameter estimates unstable if many variables were included in the same model. All models for exposures except maternal age were adjusted for potential confounders, including maternal age, race, and education level and index child's sex. The model for maternal age was adjusted for all other potential confounders listed above. Age was considered as a continuous variable, because it is plausible that risk increases in a linear fashion as age increases; a quadratic effect also was explored, because some studies have reported increased risk both for decreased age and for increased age at conception. The functional form of age as a continuous variable was examined using cumulative residuals.40 Age also was examined as a categorical variable. Subgroup analysis was performed to examine the effects of the exposures on ALL and AML separately with comparison to the entire control group. Given the exploratory nature of this analysis, no adjustment for multiple comparisons was made. All analyses were performed using SAS version 9.1 (SAS Institute Inc., Cary, NC).
Demographic information on cases and controls and their mothers is presented in Table 1. There were some differences between case and control mothers in educational attainment and race. Case mothers were more likely to be nonwhite (OR, 1.93; 95% CI, 1.05–3.54) and were less likely to have more than a high school education (OR, 0.49; 95% CI, 0.30–0.78). The French-American-British morphology subtypes of ALL and AML41 were unknown for 53% of ALL cases and 31% of AML cases. Among those who had subtype information available, 78% of ALL cases had an L1 subtype, 11% had an L2 subtype, 4% had an L1/L2 subtype, and 6.5% had some other subtype. Among AML cases who had subtype information available, 60% had an M7 subtype, 10% had an M0 subtype, 7.5% had an M1 subtype, 15% had an M2 subtype, 2.5% had an M4 subtype, and 5% had an M6 subtype. The high proportion of cases with an M7 subtype was in accordance with previous findings in the DS population.42
Table 1. Baseline Characteristics by Leukemia Subtype
| White||152 (88.4)||126 (79.7)||1.0|| ||79 (81.4)||1.0|| ||47 (77)||1.0|| |
| Nonwhite||20 (11.6)||32 (20.3)||1.93||1.05–3.54||18 (18.6)||1.73||0.87–3.46||14 (23)||2.26||1.06–4.83|
| ≤High school||41 (23.8)||62 (39.2)||1.0|| ||40 (41.2)||1.0|| ||22 (36.1)||1.0|| |
| >High school||131 (76.2)||96 (60.8)||0.49||0.30–0.78||57 (58.8)||0.45||0.26–0.76||39 (63.9)||0.56||0.30–1.04|
|Household income, $US|
| ≤30,000||57 (33.1)||57 (36.5)||1.0|| ||37 (38.5)||1.0|| ||20 (33.3)||1.0|| |
| 30,001–50,000||41 (23.8)||43 (27.6)||1.05||0.60–1.84||22 (22.9)||0.83||0.43–1.60||21 (35)||1.46||0.70–3.04|
| >50,000||74 (43)||56 (35.9)||0.76||0.46–1.25||37 (38.5)||0.77||0.44–1.36||19 (31.7)||0.73||0.36–1.50|
|Index child's sex|
| Boy||90 (52)||85 (53.8)||1.0|| ||57 (58.8)||1.0|| ||28 (45.9)||1.0|| |
| Girl||83 (48)||73 (46.2)||0.93||0.60–1.44||40 (41.2)||0.76||0.46–1.26||33 (54.1)||1.28||0.71–2.30|
|Index child's age at reference date, y*|
| <2||76 (43.9)||63 (39.9)|| || ||13 (13.4)|| || ||50 (82)|| || |
| 2–5||52 (30.1)||66 (41.8)|| || ||55 (56.7)|| || ||11 (18)|| || |
| ≥6||45 (26)||29 (18.3)|| || ||29 (29.9)|| || ||0 (0)|| || |
Table 2 presents the adjusted analysis for reproductive variables. Age as a continuous variable was a significant predictor of AML (OR, 1.06; 95% CI, 1.01–1.12) but was not a significant predictor either overall or for ALL. This association was not attenuated by ever having received treatment for infertility. There also was a significant difference in the risk of AML between mothers aged ≥35 years compared with mothers aged <30 years (OR, 2.63; 95% CI, 1.26–5.49); this association also was significant for overall leukemia (OR, 1.76; 95% CI, 1.03–3.02). In a model that included a linear and quadratic term for age, the quadratic term was not statistically significant (data not shown).
Table 2. Association Between Childhood Leukemia and Reproductive History
|Prior fetal loss|
| None||123 (71.1)||114 (72.2)||1.0|| ||73 (75.3)||1.0|| ||41 (67.2)||1.0|| |
| 1||42 (24.3)||29 (18.4)||0.67||0.38–1.19||13 (13.4)||0.50||0.25–1.03||16 (26.2)||0.98||0.48–2.01|
| ≥2||8 (4.6)||15 (9.5)||1.92||0.75–4.91||11 (11.3)||2.16||0.77–6.02||4 (6.6)||1.39||0.38–5.17|
|Time since last live birth, y|
| None or ≤2||76 (44.2)||67 (42.4)||1.0|| ||41 (42.3)||1.0|| ||26 (42.6)||1.0|| |
| From >2 to ≤4||58 (33.7)||41 (25.9)||0.82||0.48–1.41||29 (29.9)||0.99||0.53–1.83||12 (19.7)||0.58||0.26–1.26|
| >4||38 (22.1)||50 (31.7)||1.33||0.73–2.42||27 (27.8)||1.29||0.63–2.61||23 (37.7)||1.39||0.65–3.00|
|Birth order of the index child|
| First||55 (32)||47 (29.8)||1.0|| ||30 (30.9)||1.0|| ||17 (27.9)||1.0|| |
| Second||57 (33.1)||56 (35.4)||1.11||0.63–1.96||29 (29.9)||1.00||0.51–1.94||27 (44.3)||1.35||0.64–2.84|
| Third or higher||60 (34.9)||55 (34.8)||0.83||0.44–1.54||38 (39.2)||1.05||0.51–2.15||17 (27.9)||0.54||0.23–1.28|
| Mean ± SD, y†||31.9 ± 6.2||32.5 ± 7.1||1.03||1.00–1.07||31.6 ± 7.4||1.01||0.97–1.06||33.8 ± 6.5||1.06||1.01–1.12|
| Categorical y|
| <30||66 (38.4)||57 (36.1)||1.0|| ||40 (41.2)||1.0|| ||17 (27.9)||1.0|| |
| 30–34||54 (31.4)||38 (24)||0.93||0.53–1.63||23 (23.7)||0.78||0.41–1.49||15 (24.6)||1.23||0.55–2.76|
| ≥35||52 (30.2)||63 (39.9)||1.76||1.03–3.02||34 (35.1)||1.44||0.77–2.71||29 (47.5)||2.63||1.26–5.49|
Several factors were not associated with leukemia overall or by subtype, including prior fetal loss, interval since last live birth, birth order, and prior contraceptive use (Tables 2 and 3). Small numbers of exposed children prevented logistic regression analysis of previous molar pregnancy or previous ectopic pregnancy.
Table 3. Association Between Childhood Leukemia and Contraceptive Use
|Any use prior to index pregnancy|
| Oral contraceptives|
| No||35 (20.2)||44 (27.8)||1.0|| ||25 (25.8)||1.0|| ||19 (31.1)||1.0|| |
| Yes||138 (79.8)||114 (72.2)||0.66||0.39–1.13||72 (74.2)||0.75||0.41–1.39||42 (68.9)||0.58||0.29–1.17|
| Implant or injectable contraceptives|
| No||161 (93.1)||150 (94.9)||1.0|| ||93 (95.9)||1.0|| ||57 (93.4)||1.0|| |
| Yes||12 (6.9)||8 (5.1)||0.61||0.23–1.58||4 (4.1)||0.50||0.15–1.62||4 (6.6)||0.82||0.24–2.83|
|Use in the 2 y prior to index pregnancy|
| Barrier methods|
| No||86 (49.7)||77 (48.7)||1.0|| ||46 (47.4)||1.0|| ||31 (50.8)||1.0|| |
| Yes||87 (50.3)||81 (51.3)||1.04||0.66–1.61||51 (52.6)||1.11||0.67–1.85||30 (49.2)||0.98||0.54–1.80|
| No||155 (89.6)||132 (84.1)||1.0|| ||81 (83.5)||1.0|| ||51 (85)||1.0|| |
| Yes||18 (10.4)||25 (15.9)||1.38||0.71–2.71||16 (16.5)||1.48||0.70–3.15||9 (15)||1.15||0.46–2.85|
| No||167 (96.5)||154 (98.7)|| || ||95 (97.9)|| || ||59 (100)|| || |
| Yes||6 (3.5)||2 (1.3)||†|| ||2 (2.1)||†|| ||0 (0)||†|| |
| Other method|
| No||166 (96)||151 (95.6)||1.0|| ||90 (92.8)||1.0|| ||61 (100)|| || |
| Yes||7 (4)||7 (4.4)||0.97||0.32–2.98||7 (7.2)||1.73||0.56–5.38||0 (0)||†|| |
Some variables that were related to infertility were associated significantly with acute leukemia, particularly AML (Table 4). There was an increased risk of AML among children with DS whose parents had ever tried for ≥1 year to become pregnant (OR, 2.22; 95% CI, 1.14–4.33). In addition, there was some suggestion of an association between AML in DS children whose parents had undergone surgical infertility treatment (OR, 2.78, 95% CI, 0.84–9.18) or whose parents had received other infertility treatment (OR, 3.31; 95% CI, 0.81–13.57). The model that included both surgical and other infertility treatments still showed increased ORs, although they were attenuated slightly for each treatment (surgical treatment: OR, 2.38; 95% CI, 0.58–9.79; other treatments: OR, 2.07; 95% CI, 0.40–10.76). No association was observed between infertility treatment and medication or length of time to conceive the index child in the overall group or in the ALL or AML subgroups.
Table 4. Association Between Childhood Leukemia and Infertility
|Ever try for ≥1 y|
| No||134 (77.5)||115 (72.8)||1.0|| ||78 (80.4)||1.0|| ||37 (60.7)||1.0|| |
| Yes||39 (22.5)||43 (27.2)||1.32||0.78–2.23||19 (19.6)||0.81||0.43–1.55||24 (39.3)||2.22||1.14–4.33|
|Ever visit a physician or clinic for pregnancy|
| No||145 (83.8)||131 (82.9)||1.0|| ||82 (84.5)||1.0|| ||49 (80.3)||1.0|| |
| Yes||28 (16.2)||27 (17.1)||1.15||0.63–2.11||15 (15.5)||1.03||0.51–2.09||12 (19.7)||1.36||0.61–3.01|
| No||166 (96)||146 (92.4)||1.0|| ||91 (93.8)||1.0|| ||55 (90.2)||1.0|| |
| Yes||7 (4)||12 (7.6)||2.16||0.80–5.80||6 (6.2)||1.76||0.55–5.63||6 (9.8)||2.78||0.84–9.18|
| No||157 (90.8)||142 (89.9)||1.0|| ||89 (91.8)||1.0|| ||53 (86.9)||1.0|| |
| Yes||16 (9.2)||16 (10.1)||1.10||0.51–2.35||8 (8.2)||0.89||0.36–2.24||8 (13.1)||1.43||0.55–3.72|
| No||167 (97.1)||152 (96.2)||1.0|| ||95 (97.9)||1.0|| ||57 (93.4)||1.0|| |
| Yes||5 (2.9)||6 (3.8)||1.53||0.44–5.28||2 (2.1)||†|| ||4 (6.6)||3.31||0.81–13.57|
|Time to index pregnancy|
| Not trying||77 (44.8)||69 (43.7)||1.0|| ||45 (46.4)||1.0|| ||24 (39.3)||1.0|| |
| Trying ≤1 y||79 (45.9)||73 (46.2)||1.16||0.72–1.88||43 (44.3)||0.98||0.57–1.70||30 (49.2)||1.39||0.72–2.69|
| Trying >1 y||16 (9.3)||16 (10.1)||1.45||0.65–3.21||9 (9.3)||1.09||0.43–2.75||7 (11.5)||2.11||0.73–6.14|
The association between ever trying for ≥1 year to become pregnant and other infertility variables also was examined. All of those who had tried for >1 year to conceive the index pregnancy had ever tried for ≥1 year, whereas only approximately 16% of those who were not trying or who tried for ≤1 year for the index pregnancy had ever tried for ≥1 year. Of those who said that they had received infertility treatment (medication use, surgical procedure, other procedure), 92% also reported ever having tried for ≥1 year. In an ad hoc multivariate analysis that included all other infertility variables, there was no substantial change in the results from having ever tried for ≥1 year (data not shown).
Although DS confers a greatly increased risk of leukemia, the cumulative risk in DS children is only approximately 2% by age 5 years, suggesting a role for additional factors.43 We undertook the current study to investigate possible modulators of risk in a susceptible population. In this analysis, generally, we observed no associations between acute leukemia and maternal reproductive history in children with DS. However, we did observe some suggestive associations between infertility or treatment for infertility and leukemia in the AML subgroup. We also observed an association between maternal age and leukemia overall and in the AML subgroup.
We explored the association between the several different infertility variables that were examined in this analysis. Although there was some difference in point estimates and the significance of results for mothers who had ever tried for ≥1 year to conceive and those who had tried for >1 year to conceive the index pregnancy, the results were fairly concordant. Compared with children whose parents who had not tried, there was an increased but imprecise risk of AML for children whose parents who had tried for >1 year for the index pregnancy (OR, 2.11; 95% CI, 0.73–6.14), similar to the results for ever having tried for ≥1 year to conceive. Overall, the infertility variables point to a potential increase in the risk of AML because of infertility or infertility treatment.
Maternal age is among the most commonly examined risk factors related to reproductive history. Although there has been no formal meta-analysis, there does appear to be a slight increased risk of leukemia because of advanced maternal age (>35 years) in the general population. Relative risks for advanced maternal age and childhood leukemia fluctuated around 1.3 in Little's review of >25 studies published before 199811; several subsequent studies have reported similar findings.10, 16, 19, 23, 44 A small case-control study in Mexico examined at maternal age as a risk factor for acute leukemia in children with DS and reported no association.45 Age was associated with acute leukemia risk as a continuous and categorical variable. Both variables indicated higher risk with increased age overall and in the AML subgroup.
It has been reported in a few studies that prior fetal loss significantly increased the risk of acute leukemia,12, 25 whereas most others reported no difference in risk.15–17, 20–24 One study demonstrated an inverse association.14 In the current study, we did not observe an increased risk of leukemia because of prior fetal loss either overall or within the ALL or AML subgroups.
A recent report indicated that there was an increased risk of acute leukemia (both ALL and AML) with previous molar pregnancies and indicated no correlation with previous ectopic pregnancies.17 Unfortunately, we did not have enough previous molar or ectopic pregnancies to adequately examine their correlation statistically in the current analysis. Anecdotally, the percentage of either prior molar or ectopic pregnancy was higher in the case group than in the control group.
Limited evidence has been reported for an increased leukemia risk based on interval since last live birth,16, 22 birth order,12, 14, 15, 20, 29, 30 and prior contraceptive use.15, 26–28 Similarly, we observed no association with any of those variables and acute leukemia in children with DS.
Our current results indicate that infertility or infertility treatment may be related to AML in children with DS. A recent review by Lightfoot et al. suggested several different biologic mechanisms by which infertility or infertility treatment may lead to cancer, including the use of antiestrogens, which are similar to diethylstilbestrol; less restrictive gamete selection, which may result in the transmission of damaged DNA; or epigenetic defects caused by IVF culture and manipulation.46 The most notable finding in our study was a significant association of couples who had ever tried unsuccessfully for ≥1 year to conceive in the AML subgroup. The rate of infertility (defined as not using contraception and not pregnant for ≥12 months) for married women was estimated at 7.4% in the National Survey of Family Growth in 2002, a rate that was lower than what we observed in our control group (22.5% for ever unsuccessfully tried for ≥1 year to conceive). Roman et al. reported a similar magnitude of risk for AML and ALL combined among children whose mothers had ever had medical investigation of infertility (OR, 2.1; 95% CI, 0.9–4.6), who ever had received treatment for infertility (OR, 2.1; 95% CI, 0.7–6.4), and who themselves were conceived after treatment for infertility (OR, 2.0; 95% CI, 0.6–6.9).33 Most parents in the study by Roman et al. who ever had been treated were treated for the index child (5 of 7 couples for both cases and controls). Shu et al. reported no difference in risk for whether or not the index pregnancy was planned, for the time to conception, or for receiving medical help for conception; however, those authors did not report ORs, making comparison to our study difficult.15 Although our results are suggestive, interpretations are limited because of the small sample size and the lack of focus, for many questions, on the index pregnancy.
The strengths of the current study included a very specific population of children with DS, among whom there is very little research on leukemia. In addition to yielding clues to the etiology of leukemia among children with DS, this research also may offer insight into leukemia risk in the general population. To our knowledge, this is the largest case-control study of leukemia in a DS population and the first to examine many potential risk factors, including reproductive history, contraceptive use, and infertility. The cases included in this study are expected to represent a large portion of children with DS treated in the United States, because COG institutions treat an estimated 85% of patients with leukemia aged <15 years and 73% of patients with leukemia aged <20 years.47
There is a possibility of recall bias, because this was a case-control study with data collected by interview. However, the few studies that examined recall bias in maternal responses to events during and around pregnancy produced little evidence that information on reproductive history was reported differentially between cases and controls. In addition, maternal report has had generally good agreement with medical records.12, 48, 49 To our knowledge, no studies to date have examined the potential recall bias with respect to infertility problems or treatment; thus, the potential for recall bias in these variables is not known.
Another limitation of our study was the possibility of selection bias. Controls were selected from rosters that were provided by the cases' primary care physicians; thus, bias may have arisen if those physicians were not representative of all primary care physicians who treat children with DS. In addition, not all physicians provided rosters; thus, there also may have been some bias if the physicians who did and did not provide rosters had patients with different characteristics. However, when the analysis was restricted to the cases whose primary care physician provided controls, the results for significant findings did not change.
We also lacked power, particularly in subgroup analyses, because of small sample sizes. This study was exploratory and made many comparisons for which we did not correct the nominal P value; therefore, the few significant findings may have been caused by chance alone. Finally, questions about infertility problems and treatment referred mostly to the mothers' entire reproductive history and were not specific to the index child. Although this complicates interpretation, there is at least 1 study that reported consistent results for either ever receiving infertility treatment or receiving infertility treatment for the index child.33
The prevalence of conception after infertility and treatment for infertility probably is elevated among children with DS. The ability to conceive naturally decreases with age; in the United States, it is estimated that about 19% of women aged >40 years have ever received services for infertility compared with 5% of women ages 15 to 29 years.50 Currently, older maternal age is a well-known risk factor for DS.35 Chromosomal defects, among which DS is the most common, also appear to occur more often in couples who are treated for infertility compared with couples who conceive children naturally.36, 37 Given these circumstances, our findings regarding history of infertility and increased risk of AML among children with DS invite elucidation.