Milk Intake, Height and BMI in Preschool Children
The ECLS-B is a large, multisource, multimethod study sponsored by the National Center for Education Statistics (NCES—US Department of Education) to examine influences on early childhood experiences. This nationally representative sample of children born in 2001 was selected by randomly sampling >14 000 birth certificates, with a final sample of approximately 10 700 completed parent interviews (77% response rate). Parents gave informed consent. We used data from the 4-year-old and 5-year-old evaluations where information on milk consumption was gathered, enabling prospective analysis among preschoolers.
During the 4-year visit, parents were interviewed in their home by trained assessors. The primary caregiver (most often the mother) completed a computer-assisted interview. Parents were asked a set of questions regarding the type and frequency of beverage intake, including: 'During the past 7 days, how many times did your child drink milk?' Parents were instructed to include all types of milk from a glass, cup or carton, or with cereal. They were instructed that the ½-pint of milk served at school equals one glass (8 ounces, 236 mL). Categories for frequency included no intake during the past week, 1–3 times per week, 4–6 times per week, once daily, twice daily, three times daily and ≥4 times daily. For purposes of reporting prevalence data, these quantities were converted to 0, <1, 1, 2, 3 and ≥4 servings daily. In addition, parents were asked if their child usually drinks whole milk, 2%, 1%, skim, soy or other. Parents were similarly asked about the amount of sugar-sweetened beverages (SSB) their child consumed.
Direct measurements of weight were obtained by trained researchers using standardised protocols and equipment including a digital scale. Children were dressed in light clothing without shoes. BMI was calculated as weight (kilogram)/(height (meter)). Gender-specific percentiles and z-scores for BMI, height and weight-for-height (comparing a particular child's weight to standards for reference children with that exact height) were generated using SAS code from the Centers for Disease Control and Prevention growth measures. Weight categories were designated normal weight (<85th%), overweight (≤85th–<95th%) and obese (≥95th%).
Parents identified their child's gender and race/ethnicity. Race/ethnicity was grouped into five categories: white, black, Asian, Hispanic and other. NCES calculated socioeconomic status (SES) based on family income, maternal education, maternal occupation, paternal education and paternal occupation. Participants were categorised into SES quintiles (lowest SES=1; highest SES=5).
We performed all analyses using SAS software, V.9.3 (SAS Institute, Cary, North Carolina, USA), using survey procedures with sampling weights provided by the NCES to account for the complex sampling design. All statistical significance tests were two-sided, with significance of α=0.05. Unweighted sample sizes were rounded to the nearest 50 as per the NCES rules. To better compare associations of volume of milk among milk drinkers, we excluded children who did not drink milk. To assess the longitudinal links, we evaluated the association of milk intake at 4 years on anthropometry outcomes at 5 years. This approach was used previously to minimise reverse-causality. Using multivariable linear regression models, we performed both cross-sectional and longitudinal analyses. First we regressed: (i) age 4-year and 5-year z-scores for BMI, height and weight-for-height on milk consumption categories (<1, 1, 2, 3, ≥4 glasses of milk daily) at 4 years in cross-sectional analysis and (ii) longitudinal changes in these z-scores (eg, 5-year BMI z-score minus 4-year BMI z-score) on baseline milk consumption categories. Similarly, we used multivariable logistic regression models to examine the odds of overweight/obese across the milk consumption categories. We also compared the odds of consumption of SSBs among non-drinkers of milk compared with drinkers. Regression coefficients, ORs and 95% CIs are reported in the text and/or tables. All multivariable models were adjusted for sex, race/ethnicity, SES and milk type. We assessed our regression models for interactions between the potential confounder variables. Using the LSMEANS in SAS, we adjusted for multiple comparisons with a Bonferroni correction.
Methods
Data Set
The ECLS-B is a large, multisource, multimethod study sponsored by the National Center for Education Statistics (NCES—US Department of Education) to examine influences on early childhood experiences. This nationally representative sample of children born in 2001 was selected by randomly sampling >14 000 birth certificates, with a final sample of approximately 10 700 completed parent interviews (77% response rate). Parents gave informed consent. We used data from the 4-year-old and 5-year-old evaluations where information on milk consumption was gathered, enabling prospective analysis among preschoolers.
Measures
During the 4-year visit, parents were interviewed in their home by trained assessors. The primary caregiver (most often the mother) completed a computer-assisted interview. Parents were asked a set of questions regarding the type and frequency of beverage intake, including: 'During the past 7 days, how many times did your child drink milk?' Parents were instructed to include all types of milk from a glass, cup or carton, or with cereal. They were instructed that the ½-pint of milk served at school equals one glass (8 ounces, 236 mL). Categories for frequency included no intake during the past week, 1–3 times per week, 4–6 times per week, once daily, twice daily, three times daily and ≥4 times daily. For purposes of reporting prevalence data, these quantities were converted to 0, <1, 1, 2, 3 and ≥4 servings daily. In addition, parents were asked if their child usually drinks whole milk, 2%, 1%, skim, soy or other. Parents were similarly asked about the amount of sugar-sweetened beverages (SSB) their child consumed.
Direct measurements of weight were obtained by trained researchers using standardised protocols and equipment including a digital scale. Children were dressed in light clothing without shoes. BMI was calculated as weight (kilogram)/(height (meter)). Gender-specific percentiles and z-scores for BMI, height and weight-for-height (comparing a particular child's weight to standards for reference children with that exact height) were generated using SAS code from the Centers for Disease Control and Prevention growth measures. Weight categories were designated normal weight (<85th%), overweight (≤85th–<95th%) and obese (≥95th%).
Parents identified their child's gender and race/ethnicity. Race/ethnicity was grouped into five categories: white, black, Asian, Hispanic and other. NCES calculated socioeconomic status (SES) based on family income, maternal education, maternal occupation, paternal education and paternal occupation. Participants were categorised into SES quintiles (lowest SES=1; highest SES=5).
Data Analysis
We performed all analyses using SAS software, V.9.3 (SAS Institute, Cary, North Carolina, USA), using survey procedures with sampling weights provided by the NCES to account for the complex sampling design. All statistical significance tests were two-sided, with significance of α=0.05. Unweighted sample sizes were rounded to the nearest 50 as per the NCES rules. To better compare associations of volume of milk among milk drinkers, we excluded children who did not drink milk. To assess the longitudinal links, we evaluated the association of milk intake at 4 years on anthropometry outcomes at 5 years. This approach was used previously to minimise reverse-causality. Using multivariable linear regression models, we performed both cross-sectional and longitudinal analyses. First we regressed: (i) age 4-year and 5-year z-scores for BMI, height and weight-for-height on milk consumption categories (<1, 1, 2, 3, ≥4 glasses of milk daily) at 4 years in cross-sectional analysis and (ii) longitudinal changes in these z-scores (eg, 5-year BMI z-score minus 4-year BMI z-score) on baseline milk consumption categories. Similarly, we used multivariable logistic regression models to examine the odds of overweight/obese across the milk consumption categories. We also compared the odds of consumption of SSBs among non-drinkers of milk compared with drinkers. Regression coefficients, ORs and 95% CIs are reported in the text and/or tables. All multivariable models were adjusted for sex, race/ethnicity, SES and milk type. We assessed our regression models for interactions between the potential confounder variables. Using the LSMEANS in SAS, we adjusted for multiple comparisons with a Bonferroni correction.
Source...