Choline Intake During Pregnancy and Child Cognition at Age 7
Choline Intake During Pregnancy and Child Cognition at Age 7
We studied participants in Project Viva, an ongoing prospective prebirth cohort study initiated in eastern Massachusetts in 1999. Women joined the study during their first prenatal visit at Harvard Vanguard Medical Associations, a large multispecialty group practice. Eligibility criteria included fluency in English, gestational age less than 22 weeks at the first prenatal visit, and singleton pregnancy. Additional details of recruitment and retention procedures have been published elsewhere. Of the 2,128 mother-infant pairs in the cohort, we obtained first- and/or second-trimester dietary data from 1,896 women. We administered cognitive tests to 1,038 of their children who were still in follow-up at age 7 years. We excluded mother-child pairs without complete covariate data (n = 131) and children for whom English was not the primary language (n = 12), leaving 895 mother-child pairs for this study. In the primary analysis of choline intake in relation to score on the Wide Range Assessment of Memory and Learning, Second Edition (WRAML2), Design and Picture Memory subtests, there were 861 mother-child pairs with data on first-trimester choline and 808 with data on second-trimester choline (n = 890 with WRAML2 score and at least 1 choline measurement). The institutional review boards of Harvard Pilgrim Health Care, Brigham and Women's Hospital, and Beth Israel Deaconess Medical Center approved the study protocols, and all mothers provided written informed consent.
At each of the first- and second-trimester study visits, to assess intake of choline and the other methyl donors from food, we administered an approximately 130-item semiquantitative food frequency questionnaire (FFQ) modified from the well-validated instrument used in the Nurses' Health Study and other large cohort studies and further calibrated for use in pregnancy. Previous investigators have shown that similar FFQs measure choline, vitamin B12, betaine, and folate accurately compared with other dietary assessment methods and/or relevant biomarker concentrations. In the Framingham Offspring Study, which used a similar approximately 130-item FFQ, Cho et al. found an inverse dose-response association between dietary choline intake and plasma total homocysteine concentration, suggesting that the FFQ validly assessed choline at physiologically relevant levels. For the FFQ used at the first-trimester visit (mean gestational age at visit = 11.7 (standard deviation (SD), 3.2) weeks), the time referent was "during this pregnancy," that is, from the date of the last menstrual period to the assessment. We also conducted a 33-item detailed interview about use (frequency, brand/type, dosage, and timing) of nutritional supplements in early pregnancy. We calculated total first-trimester maternal intake of the nutrients examined by summing food and supplement contributions. For the FFQ used at the second-trimester visit (mean gestational age at visit = 29.1 (SD, 2.4) weeks), the time referent was "during the last 3 months." The second-trimester FFQ itself also included questions about use of nutritional supplements, which we used to calculate total second-trimester intake. We adjusted micronutrient intake for total energy intake using a residuals method.
We administered cognitive tests at the children's age 7 in-person visits and performed periodic quality assurance tests to ensure correct administration and scoring of the tests. The WRAML2 Design and Picture Memory subtests assess visuospatial memory. The Kaufman Brief Intelligence Test, Second Edition (KBIT-2), assesses verbal and nonverbal intelligence to create a composite IQ score and is reliable and valid for use in children aged 4 years or more. We also administered this test to mothers at the child's age 7 visit to assess maternal verbal and nonverbal intelligence. All test scores were age-standardized. We assessed cognition on a continuous scale.
Using a combination of questionnaires and interviews, we collected information about a range of sociodemographic factors, lifestyle habits, and medical and reproductive history. Mothers reported their educational level, smoking during pregnancy, date of birth, and parity, the father's educational level, the child's race/ethnicity, and the child's primary language. For women with missing data on smoking status, we reviewed clinical records to assess smoking during pregnancy. We used the FFQ to obtain information on maternal fish consumption and total energy intake. We previously calibrated the questionnaire against elongated n-3 fatty acids in erythrocytes and found associations between maternal fish intake during pregnancy and child cognition at age 3 years. Mothers completed the Home Observation for Measurement of the Environment (HOME) middle childhood questionnaire, which assesses maternal interactions and the home environment. The HOME score independently predicts cognitive development and has been used in previous longitudinal studies such as the National Longitudinal Study of Youth.
After examining baseline characteristics of the study population, we calculated Spearman correlation coefficients for correlations between first- and second-trimester intakes of choline and the other methyl donor nutrients.
We conducted linear regression of the cognitive tests on maternal intake of choline using quartiles of choline intake to minimize the influence of outliers. We calculated P values for trend based on the median value within each quartile of intake, using the Wald test. We assessed the bivariate association between choline and the cognitive test outcomes; then adjusted choline for vitamin B12, betaine, and folate; and finally added variables that we considered a priori to be confounders or that were associated with maternal dietary intake of methyl donors and/or child cognitive function. We present results from sequential multivariable models to illustrate the extent to which addition of covariates changed effect estimates. The final model included adjustment for maternal intake of other methyl donors, age, race/ethnicity, education, KBIT-2 score, parity, smoking, and fish and energy intakes during pregnancy, paternal education, HOME score, and child's age and sex.
We additionally considered household income, maternal prepregnancy body mass index (weight (kg)/height (m)), maternal iron intake, exercise, and alcohol consumption during pregnancy as potential confounders in the multivariable models, but none of these variables materially changed (i.e., by >10%) exposure-outcome associations, so we did not include them in the final model. We also examined the extent to which associations were modified by child sex by examining associations separately in females versus males and by including an interaction term in the final model.
We examined maternal intake of vitamin B12 and folate/folic acid from food or supplements separately to see if one had a stronger association with the cognitive test results than the other. Finally, we separately examined whether vitamin B2, vitamin B6, methionine, iron, cadmium, and zinc—other nutrients involved in the methylation pathways—were individually associated with child cognition in the fully adjusted model.
We performed all calculations in SAS, version 9.2 (SAS Institute Inc., Cary, North Carolina).
Materials and Methods
Study Sample
We studied participants in Project Viva, an ongoing prospective prebirth cohort study initiated in eastern Massachusetts in 1999. Women joined the study during their first prenatal visit at Harvard Vanguard Medical Associations, a large multispecialty group practice. Eligibility criteria included fluency in English, gestational age less than 22 weeks at the first prenatal visit, and singleton pregnancy. Additional details of recruitment and retention procedures have been published elsewhere. Of the 2,128 mother-infant pairs in the cohort, we obtained first- and/or second-trimester dietary data from 1,896 women. We administered cognitive tests to 1,038 of their children who were still in follow-up at age 7 years. We excluded mother-child pairs without complete covariate data (n = 131) and children for whom English was not the primary language (n = 12), leaving 895 mother-child pairs for this study. In the primary analysis of choline intake in relation to score on the Wide Range Assessment of Memory and Learning, Second Edition (WRAML2), Design and Picture Memory subtests, there were 861 mother-child pairs with data on first-trimester choline and 808 with data on second-trimester choline (n = 890 with WRAML2 score and at least 1 choline measurement). The institutional review boards of Harvard Pilgrim Health Care, Brigham and Women's Hospital, and Beth Israel Deaconess Medical Center approved the study protocols, and all mothers provided written informed consent.
Measurements
At each of the first- and second-trimester study visits, to assess intake of choline and the other methyl donors from food, we administered an approximately 130-item semiquantitative food frequency questionnaire (FFQ) modified from the well-validated instrument used in the Nurses' Health Study and other large cohort studies and further calibrated for use in pregnancy. Previous investigators have shown that similar FFQs measure choline, vitamin B12, betaine, and folate accurately compared with other dietary assessment methods and/or relevant biomarker concentrations. In the Framingham Offspring Study, which used a similar approximately 130-item FFQ, Cho et al. found an inverse dose-response association between dietary choline intake and plasma total homocysteine concentration, suggesting that the FFQ validly assessed choline at physiologically relevant levels. For the FFQ used at the first-trimester visit (mean gestational age at visit = 11.7 (standard deviation (SD), 3.2) weeks), the time referent was "during this pregnancy," that is, from the date of the last menstrual period to the assessment. We also conducted a 33-item detailed interview about use (frequency, brand/type, dosage, and timing) of nutritional supplements in early pregnancy. We calculated total first-trimester maternal intake of the nutrients examined by summing food and supplement contributions. For the FFQ used at the second-trimester visit (mean gestational age at visit = 29.1 (SD, 2.4) weeks), the time referent was "during the last 3 months." The second-trimester FFQ itself also included questions about use of nutritional supplements, which we used to calculate total second-trimester intake. We adjusted micronutrient intake for total energy intake using a residuals method.
We administered cognitive tests at the children's age 7 in-person visits and performed periodic quality assurance tests to ensure correct administration and scoring of the tests. The WRAML2 Design and Picture Memory subtests assess visuospatial memory. The Kaufman Brief Intelligence Test, Second Edition (KBIT-2), assesses verbal and nonverbal intelligence to create a composite IQ score and is reliable and valid for use in children aged 4 years or more. We also administered this test to mothers at the child's age 7 visit to assess maternal verbal and nonverbal intelligence. All test scores were age-standardized. We assessed cognition on a continuous scale.
Using a combination of questionnaires and interviews, we collected information about a range of sociodemographic factors, lifestyle habits, and medical and reproductive history. Mothers reported their educational level, smoking during pregnancy, date of birth, and parity, the father's educational level, the child's race/ethnicity, and the child's primary language. For women with missing data on smoking status, we reviewed clinical records to assess smoking during pregnancy. We used the FFQ to obtain information on maternal fish consumption and total energy intake. We previously calibrated the questionnaire against elongated n-3 fatty acids in erythrocytes and found associations between maternal fish intake during pregnancy and child cognition at age 3 years. Mothers completed the Home Observation for Measurement of the Environment (HOME) middle childhood questionnaire, which assesses maternal interactions and the home environment. The HOME score independently predicts cognitive development and has been used in previous longitudinal studies such as the National Longitudinal Study of Youth.
Statistical Analysis
After examining baseline characteristics of the study population, we calculated Spearman correlation coefficients for correlations between first- and second-trimester intakes of choline and the other methyl donor nutrients.
We conducted linear regression of the cognitive tests on maternal intake of choline using quartiles of choline intake to minimize the influence of outliers. We calculated P values for trend based on the median value within each quartile of intake, using the Wald test. We assessed the bivariate association between choline and the cognitive test outcomes; then adjusted choline for vitamin B12, betaine, and folate; and finally added variables that we considered a priori to be confounders or that were associated with maternal dietary intake of methyl donors and/or child cognitive function. We present results from sequential multivariable models to illustrate the extent to which addition of covariates changed effect estimates. The final model included adjustment for maternal intake of other methyl donors, age, race/ethnicity, education, KBIT-2 score, parity, smoking, and fish and energy intakes during pregnancy, paternal education, HOME score, and child's age and sex.
We additionally considered household income, maternal prepregnancy body mass index (weight (kg)/height (m)), maternal iron intake, exercise, and alcohol consumption during pregnancy as potential confounders in the multivariable models, but none of these variables materially changed (i.e., by >10%) exposure-outcome associations, so we did not include them in the final model. We also examined the extent to which associations were modified by child sex by examining associations separately in females versus males and by including an interaction term in the final model.
We examined maternal intake of vitamin B12 and folate/folic acid from food or supplements separately to see if one had a stronger association with the cognitive test results than the other. Finally, we separately examined whether vitamin B2, vitamin B6, methionine, iron, cadmium, and zinc—other nutrients involved in the methylation pathways—were individually associated with child cognition in the fully adjusted model.
We performed all calculations in SAS, version 9.2 (SAS Institute Inc., Cary, North Carolina).
