• Down syndrome;
  • vitamin supplementation;
  • leukemia;
  • children;
  • epidemiology


  1. Top of page
  2. Abstract


Periconceptional vitamin supplementation reduces the risk of neural tube defects, and possibly may reduce the risk of certain childhood malignancies, including acute lymphoblastic leukemia (ALL). Because children with Down syndrome (DS) experience a 20-fold higher risk of leukemia than the general population, the authors evaluated whether periconceptional vitamin supplementation reduced the risk of leukemia in children with DS.


From 1997 to 2002, 158 children ages birth–18 years with DS and acute leukemia (n = 61 children with acute myeloid leukemia [AML] and n = 97 children with ALL) were enrolled through the Children's Oncology Group in North America. Children with DS alone (n = 173) were identified through the cases' pediatric clinics and frequency matched to cases on age. Mothers of cases and controls completed a telephone interview that included questions regarding vitamin supplement use in the periconceptional period and after knowledge of pregnancy.


A decreased risk of leukemia was observed with vitamin supplementation in the periconceptional period (odds ratio [OR] = 0.63; 95% confidence interval [95% CI], 0.39–1.00). When stratified by leukemia type, the reduced risk was observed for ALL (OR = 0.51; 95% CI, 0.30–0.89), but not for AML (OR = 0.92; 95% CI, 0.48–1.76). Compared with vitamin use in the periconceptional period, use only after knowledge of pregnancy was associated with an increased risk of leukemia (OR = 1.61; 95% CI, 1.00–2.58). This was observed for both ALL and AML.


These data added to a growing body of evidence that suggests that periconceptional vitamin use may be protective in the development of certain childhood cancers. Cancer 2005. © 2005 American Cancer Society.

Children with Down syndrome (DS) experience nearly a 20-fold increased risk of developing leukemia compared with children without DS.1 Because trisomy 21 is also 1 of the most common acquired genetic abnormalities observed in the leukemia cells of children without DS,2 there has been speculation that genes on chromosome 21 may play a role. Despite intense research efforts, no major genes have been identified. We have speculated that leukemia in children with DS may arise from a combination of genetic susceptibility and environmental exposures,3, 4 similar to that proposed for children without DS.5 Importantly, however, because of the increased risk of leukemia in children with trisomy 21, host and environmental factors that contribute to leukemia risk may be identified more readily.

Maternal vitamin supplementation, either shortly before or after conception, has been shown to reduce the risk of neural tube defects.6, 7 A few studies have found a reduced risk of childhood acute lymphoblastic leukemia (ALL) with maternal vitamin supplement use around the time of pregnancy.8, 9 We evaluated whether periconceptional vitamin supplement use was associated with a reduced risk of leukemia in children with DS using a case group of children with leukemia and DS, and controls with only DS.


  1. Top of page
  2. Abstract

Case Identification

Children age < 20 years with DS diagnosed with incident acute leukemia between January 1997 and October 2002 were identified through the registration files of the Children's Oncology Group (COG). COG formed from the merger of the Children's Cancer Group and the Pediatric Oncology Group and consists of > 200 member institutions, which together care for an estimated 94% of patients with leukemia < 15 years and 73% of patients with leukemia < 20 years in the U.S.10, 11 Case eligibility criteria included residence in North America at the time of diagnosis, presence of a telephone in the home, and availability of an English-speaking biologic mother for interview. Institutional review boards of the University of Minnesota and participating COG institutions approved the current study.

Between 1997 and 2002, 210 potentially eligible cases were identified through 116 COG institutions participating in the study. Interviews were successfully completed for 158 (75%). Reasons for noninterview included maternal refusal (17%), physician refusal (5%), and the inability to locate the mother (3%). Of the successfully completed telephone interviews, 97 cases were diagnosed with ALL (13 were age < 2 years, 55 were ages 2–5 years, and 29 were ages 6–19 years) and 61 were diagnosed with acute myeloid leukemia (AML) (50 were age < 2 years and 11 were ages 2–5 years).

Control Identification

After completion of the telephone interview, case mothers were asked to provide the name and address of the physician responsible for their child's primary care before diagnosis of leukemia. These physicians were contacted and asked to provide us with a roster of pediatric patients with DS in their practice who had no history of leukemia. Potential controls with DS were selected randomly from the rosters and frequency matched to case subjects on age at leukemia diagnosis (birth, ages 1–3, ages 4–6, ages 7–10, ages 11–14, and ages 15–18 years). As with cases, eligible controls were required to have an available telephone in their residence and an English-speaking biologic mother available for the telephone interview.

Of 151 pediatric clinics contacted, 77 provided rosters containing 726 potential controls with DS (date of birth and gender only). Forty-seven clinics refused or were unable to provide rosters and 27 had no other age-eligible children with DS. Of these 726 potential controls, 329 were selected randomly by us based on date of birth. After our selection, the clinic did not provide a name and address for 114 of these controls for the following reasons: the families could not be located or were no longer seen by the clinic (n = 46), the families refused the clinic's request (n = 19), the clinic determined the child was not eligible (n = 18), the clinic chose not to contact the families (n = 8), or for reasons the clinics did not report (n = 23). Thus, the names and addresses from a total of 215 mothers were forwarded to the University of Minnesota. Nine (4%) potential controls were found to be ineligible. Telephone interviews were successfully completed for 173 mothers (80.5%) of the 215 potential controls with DS. The remaining mothers of potential control children with DS were not interviewed because of refusal (11%) or inability to schedule an interview (4%).

Data Collection

Mothers were interviewed using a structured, computer-assisted telephone questionnaire. Data on demographics, medical history, pregnancy and family characteristics, and personal habits (i.e., smoking and alcohol consumption) were collected along with the index child's medical history. Information regarding maternal use of vitamin supplements (as well as iron supplements specifically) included the year before pregnancy, during the index pregnancy but before knowledge of pregnancy, and after knowledge of pregnancy. Detailed clinical information for all cases with DS was available from COG institutions and cytogenetic data were collected from the primary physician of controls with DS.

Statistical Analyses

Odds ratios (OR) and 95% confidence intervals (95% CI) were used to measure the effect of maternal vitamin supplementation on leukemia risk and were estimated using unconditional logistic regression models (SAS Version 8.02 analytical software; SAS Institute Inc., Cary, NC). Analyses included all acute leukemia combined, as well as separately by acute leukemia subgroups (ALL and AML). Secondary analyses considered the age at leukemia diagnosis, and the type and number of supplements taken. Greater than 96% of women who reported taking a vitamin supplement before or during the pregnancy described either a multivitamin or a prenatal vitamin (referred henceforth as multivitamin). Only 12 women reported taking an additional folic acid supplement, 5 took a vitamin B complex, 7 took vitamin C, 4 took vitamin E, and 7 took calcium. These small numbers precluded separate analyses.

Tests for linear trend were evaluated by treating categorical variables as ordinal variables. Multivariate-adjusted analyses were performed to evaluate the potential confounding effects of various sociodemographic characteristics and presumed leukemia risk factors.


  1. Top of page
  2. Abstract

Selected characteristics of the 158 cases and 173 controls are shown in Table 1. Overall, there was little difference between cases and controls with respect to household income, morning sickness, length of gestation, previous fetal loss, cigarette smoking during pregnancy, or alcohol drinking during pregnancy. The children were similar with respect to gender and birth weight. There was a significant difference in maternal education, with 61% of cases and 76% of controls completing some education beyond high school. There was also a significantly higher proportion of Hispanic case mothers than Hispanic control mothers. Case mothers were also slightly older than control mothers at the index child's birth.

Table 1. Baseline Characteristics According to Childhood Leukemia Status
CharacteristicsCases (n = 158) (%)Controls (n = 173) (%)OR95% CI
  1. OR: unadjusted odds ratio; 95% CI: 95% confidence interval.

Mother's characteristics    
Age at index child's birth (yrs)    
 ≤ 29 (reference)57 (36.1)66 (38.4)1.00 
 30–3650 (31.6)69 (40.1)0.840.51–1.40
 ≥ 3751 (32.3)37 (21.5)1.600.92–2.77
Length of gestation (weeks)    
 ≤ 3511 (−7.0)14 (−8.1)0.760.33–1.77
 36–3858 (36.7)71 (41.3)0.790.50–1.25
 39–40 (reference)84 (53.2)81 (47.1)1.00 
 > 405 (−3.2)6 (−3.5)0.800.24–2.74
 ≤ High school (reference)62 (39.2)41 (23.8)1.00 
 > High school96 (60.8)131 (76.2)0.490.30–0.78
 White (reference)126 (79.8)152 (88.4)1.00 
 African American7 (4.4)3 (1.7)2.820.71–11.11
 Hispanic19 (12.0)7 (4.1)3.271.33–8.04
 Other6 (3.8)10 (5.8)0.720.26–2.05
Household income    
 ≤ $30,000 (reference)57 (36.5)57 (33.1)1.00 
 $30,001–$50,00043 (27.6)41 (23.8)1.050.60–1.84
 ≥ $50,00056 (35.9)74 (43.0)0.760.46–1.25
Morning sickness    
 No (reference)50 (31.7)67 (38.7)1.00 
 Yes108 (68.4)106 (61.3)1.370.87–2.15
Previous fetal loss    
 None (reference)114 (72.2)123 (71.1)1.00 
 129 (18.4)42 (24.3)0.750.44–1.28
 ≥ 215 (9.5)8 (4.6)2.020.83–4.95
Smoking during pregnancy    
 No (reference)131 (82.9)149 (86.1)1.00 
 Yes27 (17.1)24 (13.9)1.280.70–2.33
Drinking during pregnancy    
 No (reference)131 (82.9)143 (82.7)1.00 
 Yes27 (17.1)30 (17.3)0.980.56–1.74
Index child's characteristics    
 Male (reference)85 (53.8)90 (52.0)1.00 
 Female73 (46.2)83 (48.0)0.930.60–1.44
Birth weight (g)    
 < 250022 (14.1)27 (15.6)0.880.48–1.63
 2500–4000 (reference)130 (83.3)141 (81.5)1.00 
 > 40004 (2.6)5 (2.9)0.870.23–3.30

As results were nearly identical for vitamin supplementation reported in the year before pregnancy and during early pregnancy but before knowledge of pregnancy, these time periods were combined (referred henceforth as the periconceptional period). Overall, there was a reduced risk of leukemia with any use of vitamins in the periconceptional period (OR = 0.63; 95% CI, 0.39–1.00) (Table 2). When the data were stratified on leukemia type, the reduced risk was noted for ALL (OR = 0.51; 95% CI, 0.30–0.89), but not for AML (OR = 0.92; 95% CI, 0.48–1.76), although the number of cases with AML was relatively small. In contrast, risk appeared elevated with use only after knowledge of the pregnancy (OR = 1.61; 95% CI, 1.00–2.58). This risk elevation was observed for both leukemia types. There was no obvious dose response associated with vitamin intake during either period. There was no evidence of interaction by advanced maternal age (age ≥ 35 years vs. < 35 years).

Table 2. Odds of Childhood Leukemia Associated with Maternal Vitamin Use around the Time of Pregnancy
Timing of vitamin useNo. of controls (%)No. of combined cases (%)ORa95% CIAll (%)OR95% CIAML (%)OR95% CI
  • OR: odds ratio; 95% CI: 95% confidence interval; ALL: acute lymphoblastic leukemia; AML: acute myeloid leukemia; yr: year.

  • a

    Logistic regression models were adjusted for mother's age at index pregnancy (continuous), race (white, black, Hispanic, other), and education (≤ high school vs. > high school).

  • b

    Year before and during pregnancy but before knowledge of pregnancy.

Periconceptional useb          
 No63 (36.4)76 (48.1)1.0 52 (53.6)1.0 24 (39.3)1.0 
 Yes110 (63.6)82 (51.9)0.630.39–1.0045 (46.4)0.510.30–0.8937 (60.7)0.920.48–1.76
Only after knowledge of pregnancy          
 No118 (68.2)92 (58.2)1.0 54 (55.7)1.0 38 (62.3)1.0 
 Yes55 (31.8)66 (41.8)1.611.00–2.5843 (44.3)1.700.98–2.9223 (37.7)1.450.75–2.78
Year before and throughout pregnancy          
 No67 (38.7)79 (50.0)1.0 55 (56.7)1.0 24 (39.3)1.0 
 Yes106 (61.3)79 (50.0)0.630.40–1.0142 (43.3)0.490.29–0.8437 (60.7)0.980.51–1.86
No. of vitamins taken during yr before pregnancy          
 066 (38.2)78 (49.4)1.0 52 (53.6)1.0 26 (42.6)1.0 
 191 (52.6)67 (42.4)0.630.39–1.0338 (39.2)0.560.32–0.9829 (47.5)0.830.43–1.60
 216 (9.3)13 (8.2)0.690.30–1.597 (7.2)0.550.20–1.496 (9.8)1.070.36–3.20

Additional analyses were performed to explore associations with specific age groups, recognizing these estimates are less precise due to smaller numbers. For ALL, the reduction in risk associated with periconceptional vitamin use was observed across all 3 age groups examined (age < 2 years, ages 2–5 years, and age ≥ 6 years) (data not shown). For AML, there was a reduced risk among the youngest group (OR = 0.48, 95% CI, 0.21–1.07; age < 2 years) but there was no association with the older group (OR = 1.02; 95% CI, 0.21–5.04; ages 2–5 years). The positive association with vitamin supplementation initiated after knowledge of pregnancy was most apparent in cases age < 2 years with ALL (OR = 4.27; 95% CI, 1.03–17.79), compared with cases ages 2–5 years (OR = 1.37; 95% CI, 0.58–3.24) and cases age ≥ 6 years with ALL (OR = 1.86; 95% CI, 0.65–5.29). Similar to ALL, the increased risk of AML associated with vitamin supplementation initiated after knowledge of pregnancy was confined mostly to the youngest age group (OR = 2.56; 95% CI, 1.12–5.85% [age < 2 years] vs. OR = 1.63; 95% CI, 0.32–8.41% [ages 2–5 years]). Restriction of analyses to only cases with an available DS clinic control did not alter any of these results.

Finally, we evaluated iron supplements taken in addition to that found in a multivitamin, as this was queried separately. We found no notable associations with leukemia overall, ALL, or AML for any of the time periods examined (data not shown).


  1. Top of page
  2. Abstract

Multivitamin supplementation around the time of pregnancy has been associated consistently with a reduction in birth defects, particularly oral clefts and neural tube defects.7, 12–14 These inverse associations are due primarily to folate.15 Recent studies have indicated that periconceptional vitamin supplementation may protect against certain childhood malignancies, including ALL,8, 9 neuroblastoma,16, 17 and brain tumors.18, 19 One study implicated folate specifically.9 In our study, we found a reduced risk of ALL in children with DS whose mothers reported taking multivitamin supplements either in the year before or during early pregnancy. Although this reduced risk was not observed for AML overall, there was a suggestion that maternal vitamin supplementation before knowledge of pregnancy may also reduce the risk of AML in younger children.

It is unknown why periconceptional vitamin supplementation would be associated with a decreased risk of childhood ALL. We can speculate that this protective effect may occur through antioxidant pathways or through enhancement of immune function.20 In particular, folate can influence DNA methylation as well as enhance DNA synthesis and repair,21–23 and functional polymorphisms in folate-related genes have been under scrutiny in both childhood leukemia and DS. For example, ≥ 3 studies have reported a decreased risk of ALL in children who possess the methylenetetrahydrofolate reductase (MTHFR) T allele at nucleotide 677.24–26 The MTHFR C677T genotype in the mother and the risk of bearing a child with DS also has been explored, although these data are not consistent.27–29 In the current study, it was impossible to determine whether folate played a specific role because few women reported taking folate supplements separate from what was present in a multivitamin. Unfortunately, DNA specimens were not collected from either mothers or children to evaluate genotype. Vitamin use could also be a surrogate marker for other factors that may be associated with childhood leukemia,30 including health behaviors that may not have been addressed adequately in our analysis. However, because we observed a protective overall effect for ALL and not AML, it seems unlikely that residual confounding explains these results.

Unexpectedly, we found an increased risk of leukemia (both ALL and AML) associated with initiation of vitamin supplementation only after knowledge of pregnancy. There is concern that this association may also reflect residual confounding. It is important to note that, instead of trimester-specific questions, we queried vitamin use during two pregnancy periods: before knowledge and after knowledge of pregnancy. These pregnancy periods undoubtedly could vary among women. In secondary analyses, we found no difference between cases and controls with respect to the number of weeks pregnant before confirmation by a physician. However, women who initiated vitamin use earlier in pregnancy tended to be older, more educated, have a higher income, and to be more likely to have the pregnancy confirmed by a physician within 1 or 2 months after conception. Even with adjustment for these variables, however, the increased risk remained. Although we cannot rule out other potential confounders, the positive association was most apparent for the youngest age group, which suggests that confounding might not explain these differences. Leukemias diagnosed in the first few years of life are likely initiated in utero.31 It is possible that an increased risk associated with vitamin supplementation after knowledge of pregnancy may be reflective of underlying biologic processes involving selection of certain cells, as speculated by Potter32 when reporting the adverse cancer outcomes observed in some vitamin chemoprevention trials. Some antioxidants can influence genes involved in the inhibition of apoptosis and, in turn, can enhance their effects.33–35 In the example of childhood leukemia, if the leukemia is already initiated early in pregnancy, subsequent vitamin exposure might provide a proliferative advantage to mutated cells. Of the studies that have explored vitamin supplementation and risk of childhood leukemia, none that we are aware of examined risk associated with use only after knowledge of pregnancy. This may be an area for further investigation.

The limitations of the current study need to be addressed. Although this is the largest epidemiologic study to date of DS and leukemia, the sample sizes are still small, which resulted in some imprecise effect estimates. Although response rates were similar between cases and controls, there is still the concern that not all primary care clinics provided controls with DS nor did all selected controls with DS participate. However, when we restricted the analysis to cases who had an available control with DS, our findings did not change. There is also the possibility that enrolled cases could differ from the population of children with DS in North America. However, COG-affiliated institutions treat > 90% of children with leukemia in the U.S.10, 11 Therefore, it is unlikely that a child with DS diagnosed with leukemia at any clinic in the U.S. would not be referred to and treated at a COG institution. Maternal vitamin supplementation was collected by self-report, which is subject to some inaccuracies as well as recall bias. However, a validation study that compared self-report of vitamin use with direct verification from labels of multivitamin supplement bottles reported excellent agreement (kappa = 0.90).36 Finally, it should be acknowledged that multivitamin supplements consist of various combinations and mixtures with varying degrees of potencies, which could not be evaluated in the current study.

Overall, our data are encouraging and suggest that vitamin supplementation early in pregnancy may reduce the risk of ALL in children with DS. These data add to a growing body of evidence that suggests that periconceptional vitamin use may be protective in the development of certain childhood cancers.


  1. Top of page
  2. Abstract