UC Berkeley

Acute leukemia is the most common type of cancer that occurs in children. Because many cases of childhood leukemia originate in utero, it has been hypothesized that maternal nutrition before or during pregnancy may influence the development of leukemia through its role in fetal development. Findings from studies examining the association of childhood leukemia with maternal diet before and during pregnancy have been inconsistent. Because dietary factors are known to influence the genetic and epigenetic processes involved in the development of childhood cancer, and because diet is a modifiable risk factor, maternal prenatal nutrition continues to be of interest in the etiology of childhood leukemia.

This study examined the association between maternal diet and vitamin supplement use before and during pregnancy as assessed by a food frequency questionnaire in a large and ethnically-diverse population-based case-control study, the California Childhood Leukemia Study (CCLS). Analyses were conducted with original dietary data, and employed principal components (PC) analysis to account for the high correlations between nutrients. Higher maternal intake of one-carbon metabolism nutrients from food and supplements was associated with a reduced risk of acute lymphoblastic leukemia (ALL) (odds ratio for the principal component (ORPC) = 0.91, confidence interval (CI) 0.84-0.99) and possibly acute myeloid leukemia (AML) (ORPC = 0.83, CI 0.66-1.04). The association of ALL with nutrient intake from food only was similar to the association of total nutrient intake from food and supplements, both in the study population overall and within racial/ethnic groups. However, intake of B vitamins from supplements (any versus none) was associated with a reduced risk of ALL in children of Hispanic women (OR=0.51, CI 0.28-0.94), but not in children of non-Hispanic white women (OR = 0.91, CI 0.64-1.31) or Asian women (OR = 2.24, CI 0.91-5.51).

The possible influence of exposure misclassification and selection bias on the associations observed between maternal prenatal use of vitamin supplements and risk of childhood leukemia was qualitatively and quantitatively assessed. Case and control participation and exposure assessment was reviewed in twelve published studies examining the association of childhood leukemia with maternal vitamin supplement use. The review found that most studies had low control participation, and controls generally had higher socioeconomic status than participating cases. Additionally, half of the included studies examined broad categories of vitamin supplements (e.g. prenatal vitamins) and few asked about brand or frequency of consumption. The quantitative bias analysis conducted in the CCLS data indicated that, under the assumed bias parameters, selection bias and exposure misclassification were unlikely to account for the association of vitamin supplement use and ALL observed in CCLS Hispanic women. The measures of association corrected for these systematic errors among non-Hispanic white women varied widely. However, all exposure misclassification corrections for higher sensitivity in cases (hypothesized to be the more common scenario in case-control studies) produced negative associations further away from the null value than the uncorrected estimate. In conclusion, relatively modest differences in systematic errors between Hispanic and white women could account for the heterogeneity observed in the association between vitamin supplement use and ALL by maternal ethnicity.

In addition, examination of the relationship between overall maternal diet quality, summarized by a diet quality index, and risk of childhood leukemia found that higher maternal diet quality score was associated with a reduced risk of childhood ALL (OR = 0.88, CI 0.78-0.98 for each five point increase), with a more pronounced reduction in risk observed among younger children and children of women who did not use vitamin supplements before pregnancy. There was a similar reduced risk of AML with increasing maternal diet quality score, although the confidence interval include 1.0 (OR = 0.76, CI 0.52-1.11). No single diet quality index component (i.e. food group or nutrient) appeared to account for the results, suggesting that the quality of the whole diet and the cumulative effects of many dietary components may be important in influencing childhood leukemia risk.