Iron Depletion Is Associated With Daytime Bottle-feeding in the Second and Third Years of Life

Objective  To measure the association between daytime bottle-feeding and iron depletion in young children.

Design  Cross-sectional design with concurrent measurement of exposure and outcome. The exposure was the current container (bottle or cup) used for daytime milk consumption. Child, maternal, and dietary variables were collected.

Setting  Community-based pediatric practice serving a diverse population in an urban Canadian city.

Participants  One hundred fifty healthy children, aged 12 to 38 months, attending a well-child care visit.

Main Outcome Measure  Iron depletion (serum ferritin level, <10 μg/L]).

Results  Of the 150 children, 82 (55%) were bottle-fed and 68 (45%) were cup fed. Iron depletion occurred in 29 (37%) of 78 bottle-fed and in 12 (18%) of 67 cup-fed children. The crude relative risk for iron depletion was 1.81 (95% confidence interval, 1.09-3.01). In the final logistic regression model, a significant association between bottle use and iron depletion was identified, beginning after the age of 16 months. At 18 months, the relative risk, adjusted for several child, maternal, and dietary variables, for the association between bottle use and iron depletion was 1.31 (95% confidence interval,1.24-1.47); at 24 months, the adjusted relative risk was 2.50 (95% confidence interval, 2.46-2.53). Milk consumption of more than 16 oz/d occurred in 55 (67%) of the 82 bottle-fed and in 22 (32%) of the 68 cup-fed children (P<.001).

Conclusions  In the second and third years of life, there is an almost 2-fold association between iron depletion and daytime bottle-feeding compared with cup feeding. The child's age may be a modifier, and milk volume consumed may be a mediator, of this association. Duration of bottle use is a potentially modifiable practice.

Children in the second and third years of life are vulnerable to iron depletion, which may progress to iron deficiency anemia (IDA).¹ Infants with iron deficiency are at risk for developmental delay and poor academic performance, which may not improve with iron therapy.^2-5 Iron depletion occurs in up to 30% of young children in developed countries.^6-8

Data from the 1991 US National Health Interview Survey indicate that about 95% of children aged 6 months to 5 years have ever used a baby bottle.⁹ Bottle weaning is recommended between the ages of 12 and 18 months.^10,11 However, families and physicians demonstrate considerable variation regarding the age of bottle weaning, and single-site studies^12-15 suggest that 40% of children aged 24 months and 18% of children aged 42 months continue bottle-feeding. Previous studies^16-19 suggest that prolonged bottle-feeding may contribute to iron depletion; however, the contribution of volume of milk consumed and other dietary variables were not considered.

Our study objective was to examine the association between daytime bottle-feeding and iron depletion in healthy young children in the second and third years of life, assessing the role of the child's age and volume of milk consumed, while controlling for maternal and other dietary variables.

Children were recruited from a 2-physician (S.J. and M.P.) community-based pediatric practice in Toronto, from November 1, 2002, to March 31, 2004. The practice is hospital and university affiliated, and provides primary care for approximately 10 000 children, with approximately 500 office visits per week.

Healthy children, aged 12 to 38 months, drinking nonfortified milk, were included. Exclusion criteria were as follows: acute febrile illness at enrollment, which may increase serum ferritin level; history of illness or medication associated with iron deficiency or anemia; previous diagnosis of iron depletion or anemia; current use of iron supplements; and breast milk or formula as the primary source of milk at or within 8 weeks of enrollment, which provides more available iron.

The office computerized appointment book was used to identify age-eligible children with a scheduled well-child care visit. An information letter was sent to parents 2 weeks before their scheduled appointment, informing them that the study objective was to assess the association between toddler feeding habits and iron depletion, without specific reference to bottle use to minimize reporting bias. At the appointment, the attending pediatrician and study nurse (S.W.) identified those children meeting all eligibility criteria, and asked parents to participate. Informed consent was obtained from the child's parent.

At enrollment, using a standardized data collection form, the study nurse collected data on child and maternal characteristics, child's dietary history during the first year of life, child's current use of a bottle or cup, and current diet (see details later). The study was approved by the Research Ethics Board of the Hospital for Sick Children.

Child characteristics collected included the following: age, sex, birth weight, prematurity (<37 weeks' gestation), and current weight. Maternal characteristics collected included the following: age and achievement of high school–level education or greater. The child's dietary history during the first year of life was collected as dichotomous variables (yes or no) and included the following: ever breastfed, introduction of solids at or before the age of 6 months, and introduction of cow's milk at or after the age of 12 months.

To collect data on current diet, the study team developed 8 diet questions based on the Canadian nutrition guidelines for healthy infants.²⁰ These questions quantified consumption of the following variables: milk; pure fruit juice; fruit drinks; cereals or breads; meat, fish, poultry, or legumes; red meat; vegetables and fruit; and chips, fried snacks, or sweets.

The definition of exposure incorporated time of day and type of liquid. Parents were asked the following: Has your child been completely weaned yet from daytime bottle drinking of milk? Children whose parents responded “no” were categorized as bottle-fed, and children whose parents responded “yes” were categorized as cup fed. No attempt was made to further categorize children according to nighttime bottle use or according to type of cup used.

Phlebotomy was completed in a single laboratory concurrently with determination of exposure. A venous blood sample was drawn for measurements of serum ferritin level, mean corpuscular volume, and hemoglobin level. The primary outcome, iron depletion, was defined as a serum ferritin level of less than 10 μg/L. It was anticipated that IDA would be much less prevalent than iron depletion alone, but an important outcome to explore. Iron deficiency anemia was defined as a serum ferritin level of less than 10 μg/L, a mean corpuscular volume of less than 73 μm³, and a hemoglobin level of less than 11 g/dL. These cutoff values were derived by an expert panel for the Second and Third National Health and Nutrition Examination Surveys.^21,22

The sample size calculation was based on the following estimates: α error = .05; β error = .20 (80% power); prevalence of iron depletion in the bottle-fed group, 35%; and prevalence of iron depletion in the cup-fed group, 15%. Based on these assumptions, at least 70 subjects were required per group.

The characteristics of the bottle- and cup-fed groups were compared using the Fisher exact test and the Mann-Whitney test for categorical and continuous variables, respectively. A relative risk (RR) and 95% confidence interval (CI) were calculated for iron depletion (serum ferritin level, <10 vs ≥10 μg/L).

Two distinct logistic regression models were developed. The first step was the development of a logistic regression model, predicting bottle or cup use, to calculate a propensity score for each subject. The propensity score for an individual is defined as the conditional probability of receiving the treatment given the individual's covariates.²³ The score can balance the covariates in the 2 groups and, thus, reduce bias, which can occur in observational studies. The score can then be included as a predicted probability in a subsequent logistic regression model predicting outcome. In this analysis, the propensity score logistic regression model contained variables found significant at the univariate level, to model the propensity of a child to use the bottle. Four variables were included in the score: maternal educational level, ever breastfed, daily volume of pure fruit juice consumed, and weekly servings of chips, fried snacks, or sweets. The propensity score was subsequently used as a covariate in the final model predicting iron depletion.

The second step was the development of a logistic regression model predicting iron depletion (ferritin level, <10 μg/L). The model contained 4 variables potentially related to iron depletion: bottle or cup use, the propensity score, child's age, and daily volume of milk consumed. Given the sample size, no further variables were entered into the final model. This model was examined for interaction and collinearity.

Child's age was entered into the final model rather than the propensity score to explore age as a moderator of the effect of bottle or cup use on iron depletion. A moderator variable is an effect modifier, which specifies on whom or under what conditions another variable will operate to produce the outcome.²⁴ Daily volume of milk consumed was entered into the final model to explore milk volume as a mediator of the effect of bottle use on iron depletion. A mediator variable is one that occurs in a causal pathway, causes variation in the outcome variable, and itself is caused to vary by the exposure variable.

A 2-sided P=.05 indicated statistical significance. All statistical analyses were performed using SAS statistical software, version 9.1 (SAS Institute Inc, Cary, NC).

A total of 458 children met the age criteria, of whom 332 (72%) met the full eligibility criteria. Of these children, 150 had questionnaire and blood work completed and were enrolled in the study. Maternal age ranged from 20 to 48 years, secondary education was the highest level of educational attainment for 22% of mothers, and 45% of child participants had at least 1 older sibling. Approximately 50% of mothers reported their ethnicity as other than Canadian or European, including Asian (13%), Latin American (8%), African (3%), Middle Eastern (3%), Mediterranean (2%), East Indian (1%), West Indian (1%), and First Nations (1%).

Of the 150 subjects, 82 (55%) were drinking milk from a bottle and 68 (45%) were drinking milk from a cup. The characteristics of bottle- and cup-fed children are found in Table 1 and Table 2. There were no significant differences between the bottle- and cup-fed children for sex, birth weight, prematurity, current weight, maternal age, introduction of solids by the age of 6 months, introduction of cow's milk at or after the age of 12 months, and current consumption of fruit drinks, red meat, vegetables and fruit, cereals or breads, and milk, fish, poultry, or legumes. The characteristics that were significantly different between bottle- and cup-fed children included age, secondary school as the highest level of maternal education, ever breastfed in the first year of life, and current consumption of milk, pure juice, and chips, fried snacks, or sweets.

Because of insufficient blood quantities, iron status could not be determined in 5 children (4 bottle-fed and 1 cup fed) and IDA could not be determined in 10 children (4 bottle-fed and 6 cup fed). Iron depletion (n = 145) occurred in 29 (37%) of 78 bottle-fed children and in 12 (18%) of 67 cup-fed children (RR, 1.81; 95% CI, 1.09-3.01). Iron deficiency anemia (n = 140) occurred in 3 (4%) of 78 bottle-fed children and in 1 (2%) of 62 cup-fed children (P=.40).

The propensity score logistic regression model contained variables found significant at the univariate level: maternal educational level, ever breastfed, daily volume of pure fruit juice consumed, and weekly servings of chips, fried snacks, or sweets.

The final logistic regression model contained the propensity score, current bottle use, child's age, and daily volume of milk consumed. No significant association was found between iron depletion and the propensity score (RR, 0.68; 95% CI, 0.10-2.83). Significant statistical interaction between bottle use and child's age was identified, with significant association between bottle use and iron depletion beginning after the age of 16 months (Hosmer-Lemeshow goodness-of-fit test, P=.59). The relationship between bottle use, child's age, and iron depletion is demonstrated in Figure 1. The bottle × age interaction variable was included in the model (RR, 1.13; 95% CI, 1.02-1.28), indicating that, on average, for every 1-month increase in age, bottle-fed children were 13% more likely to be iron depleted compared with cup-fed children. For example, at the age of 18 months, the RR was 1.31 (95% CI, 1.24-1.47) for the association between bottle use and iron depletion; at the age of 24 months, the RR was 2.50 (95% CI, 2.46-2.53).

In the final logistic regression model, there was no significant association between iron depletion and daily volume of milk consumed (RR, 1.03; 95% CI, 0.87-1.22). There was no evidence of interaction or collinearity when the variable measuring daily milk volume consumed was entered into the final logistic regression model. Further exploratory analysis identified that bottle or cup use was significantly associated with daily volume of milk consumed (Table 2 and Figure 2); 55 (67%) of 82 bottle-fed children and 22 (32%) of 68 cup-fed children consumed more than 16 oz/d (P<.001). At all volumes of milk consumed, bottle-fed children had an increased probability of iron depletion compared with cup-fed children (Figure 3).

In this sample of healthy children, aged 12 to 38 months, 55% were bottle-fed in the daytime, 51% consumed greater than 16 oz/d of milk, 28% had iron depletion, and 3% had IDA. Children drinking milk from a bottle during the daytime were almost twice as likely to be iron depleted compared with cup-fed children. The association remained after analyses accounted for several child characteristics, diet variables, and maternal characteristics. Age seems to be a moderator, with children older than 16 months demonstrating the strongest association between bottle use and iron depletion. Daily volume of milk consumed may act as a mediator, with bottle-fed children consuming larger volumes of milk per day and having a higher probability of iron depletion at all volumes of milk, compared with cup-fed children.

Healthy children were enrolled from an urban community primary care practice and demonstrated socioeconomic diversity. Eligibility criteria ensured that children with conditions or variables known to be strongly associated with the exposure or outcome were excluded. Accuracy of the outcome measurement was ensured by performing a standardized biochemical laboratory test (measuring serum ferritin level) at enrollment and using an accepted cutoff of less than 10 μg/L. Concurrent measurement of exposure and the primary outcome decreased the potential for recall bias. To balance covariates in the bottle and cup groups, a score measuring the child's propensity for bottle use was developed and entered into the final regression model, thus minimizing confounding bias.

Study limitations include recruitment from a single primary care practice, assessment of iron depletion rather than IDA (which required a larger sample size) as the primary outcome, and use of parental report of diet variables. Diet questions, individually or in combination, do not have adequate sensitivity and specificity to identify IDA.^25,26 The question used to classify exposure status did not allow us to examine the role of mixed bottle-feeding and cup feeding. However, the question was designed to avoid misclassification of children who were predominately bottle users.

The findings in the present study share some similarities with several studies conducted in urban US centers. Graham et al¹⁶ found that children of Southeast Asian ethnicity were more likely to be iron deficient than children of other ethnicities (12 of 17 vs 2 of 21 children), and were more likely to be bottle-fed at the age of 2 years (17 of 17 vs 10 of 21 children). These researchers used the zinc protoporphyrin–heme ratio as a measure of iron deficiency (identified to be altered by other causes, such as lead poisoning, febrile illness, or some hemoglobinopathies). Lampe and Velez¹⁷ found that, among thirty-four 18-month-old children from a primary care practice, bottle-feeding was associated with significantly greater daily milk consumption (24 vs 16 oz; P<.001) and a lower ferritin concentration (18.6 vs 16.4 μg/L; P value not significant). The small sample size may have precluded a finding of significance. Bonuck and Kahn¹⁸ found that, among 95 inner-city Hispanic and African American low-income children aged 18 months to 5 years, bottle-feeding, compared with not bottle-feeding, was associated with anemia (15 of 61 vs 5 of 34 children; P<.05). The researchers defined IDA using hemoglobin level and hematocrit alone, without measures of iron stores. Brotanek et al,¹⁹ using cross-sectional data from a US national health survey, found that bottle-feeding to the age of 24 to 48 months and Mexican ethnicity were associated with iron deficiency (odds ratio, 2.8 and 2.9, respectively). In all of these studies, the contributions of milk volume and other diet variables were not available or analyzed.

Other investigators have identified an association between volume of milk consumed and iron status. Cowin et al²⁷ found that ferritin levels were negatively associated with the amount of cow's milk consumed in 18-month-old children. Lozoff et al²⁸ found that children with IDA, aged 12 to 23 months, consumed more milk per day and took more bottles. This supports the concept that milk volume consumed may be a mediator in the causal pathway, causing variation in iron status (outcome) and itself caused to vary by bottle use (exposure).

Risk factors and mechanisms for iron depletion in the second and third years of life may differ from those in the first year of life. James et al²⁹ found poor correlation between children with IDA at the end of the first year (14 months) and children with IDA at the end of the second year (24 months). Lind et al³⁰ have suggested that dietary iron may be channeled to erythropoiesis in the first year of life and to storage in the second year of life. In developed countries, promotion of breastfeeding and iron-fortified cereals and formula and delayed introduction of whole cow's milk have contributed to the primary prevention of iron depletion in the first year of life. Less is known about the primary prevention of iron depletion in the second and third years of life.

Secondary prevention through population screening of children in the second year of life has been controversial because of lack of clear evidence of the benefits of iron therapy in reversing the developmental outcomes of children with IDA,⁵ the changing patterns of IDA from the first to the second year of life,²⁹ and the lack of evidence regarding the feasibility of screening.³¹ Identifying risk factors for iron depletion for children in the second and third years of life might lead to the development of secondary prevention strategies for high-risk groups and primary prevention strategies for the population.

There are several possible mechanisms by which bottle-feeding in 1- to 3-year-old children might lead to iron depletion. The bottle may act as a vehicle for excessive milk consumption, which may compromise iron absorption or the intake of iron-rich foods or juices. In this study, bottle-fed children consumed larger volumes of milk and smaller amounts of pure juice compared with cup-fed children.

The bottle may interfere with the development of age-appropriate feeding practices or may be associated with behavioral and developmental feeding difficulties. Previous research²⁹ has shown that children described as difficult to feed at the age of 2 years had a risk ratio of 3.5 (95% CI, 1.9-6.7) for anemia. The interaction between diet and developmentally appropriate feeding practices, and their association with iron depletion, is likely to be complex.

Results from this and other studies suggest that prolonged bottle-feeding may be a risk factor for iron depletion in children in the second and third years of life. Children using a bottle beyond the age of 12 to 16 months may be selected for screening and counseling. Bottle use is a potentially modifiable practice. The promotion of bottle weaning before the age of 12 months, or weaning directly from the breast to the cup, may be a potentially simple primary prevention strategy. These hypotheses must be tested in prospective trials.

Correspondence: Patricia C. Parkin, MD, FRCPC, Division of Pediatric Medicine and Pediatric Outcomes Research Team, Hospital for Sick Children, 555 University Ave, Toronto, Ontario, Canada M5G 1X8 (patricia.parkin@sickkids.ca).

Accepted for Publication: June 5, 2006.

Author Contributions: The authors completed all aspects of the study (design, conduct, data collection, management, analysis, interpretation, manuscript preparation, and approval) independent of the funding organization. Study concept and design: Sutcliffe, Khambalia, Jacobson, Peer, and Parkin. Acquisition of data: Sutcliffe, Khambalia, Westergard, Jacobson, Peer, and Parkin. Analysis and interpretation of data: Sutcliffe, Khambalia, and Parkin. Drafting of the manuscript: Sutcliffe, Khambalia, and Parkin. Critical revision of the manuscript for important intellectual content: Sutcliffe, Khambalia, Jacobson, Peer, and Parkin. Statistical analysis: Khambalia and Parkin. Obtained funding: Sutcliffe, Khambalia, and Parkin. Administrative, technical, and material support: Khambalia, Jacobson, Peer, and Parkin. Study supervision: Khambalia and Parkin.

Financial Disclosure: None reported.

Funding/Support: This study was supported by grant-in-aid-01 from Danone Institute of Canada. The Pediatric Outcomes Research Team is supported by a grant from the Hospital for Sick Children Foundation.

Role of the Sponsor: The funding bodies had no role in data extraction and analyses, in the writing of the manuscript, or in the decision to submit the manuscript for publication.

Acknowledgments: We thank Derek Stephens, MSc, and Cristina Goia, MSc, biostatisticians who helped in the analysis and interpretation of data; Gurvinder Dohl for patient recruitment in the first part of the study; and Nike Onabajo for assistance in the preparation of the manuscript.