Video Abstract
Addressing food insecurity while promoting healthy body weights among children is a major public health challenge. Our objective is to examine longitudinal associations between food insecurity and obesity in US children aged 1 to 19 years.
Sources for this research include PubMed, CINAHL, and Scopus databases (January 2000 to February 2022). We included English language studies that examined food insecurity as a predictor of obesity or increased weight gain. We excluded studies outside the United States and those that only considered the unadjusted relationship between food security and obesity. Characteristics extracted included study design, demographics, methods of food security assessment, and anthropometric outcomes.
Literature searches identified 2272 articles; 13 met our inclusion criteria. Five studies investigated the relationship between food insecurity and obesity directly, whereas 12 examined its relationship with body mass index or body mass index z-score. Three studies assessed multiple outcomes. Overall, evidence of associations between food insecurity and obesity was mixed. There is evidence for possible associations between food insecurity and obesity or greater weight gain in early childhood, for girls, and for children experiencing food insecurity at multiple time points. Heterogeneity in study methods limited comparison across studies.
Evidence is stronger for associations between food insecurity and obesity among specific subgroups than for children overall. Deeper understanding of the nuances of this relationship is critically needed to effectively intervene against childhood obesity.
The United States faces 2 important public health challenges in reducing childhood obesity while ensuring that children and their families have enough nutritious food for an active, healthy life. From 2017 to 2018, ∼20% of US children aged 2 to 19 were estimated to have obesity, a prevalence level that has increased by nearly 40% over the past 20 years.1 In 2020, the first year of the COVID-19 pandemic, food insecurity among US households with children was 14.8%, an increase over the 2019 level of 13.6% and a reversal of the declining trend observed over the previous decade.2
Childhood obesity and food insecurity are more prevalent in lower-income households,2,3 suggesting a potentially simultaneous occurrence of both under- and over-nutrition. Despite almost 3 decades since these dual problems were first observed,4 no consensus exists about the underlying mechanisms of their relationship. The increases in food insecurity2,5 and accelerated weight gain6–8 observed among children during the COVID-19 pandemic indicate that greater understanding of how these 2 issues interact is of great importance for child health, particularly in terms of associations between food insecurity episodes and weight status over the long term.
In this systematic review, we examined longitudinal associations between food insecurity and obesity in US children aged 1 to 19 years. The review summarizes the overall evidence, then discusses differences in the evidence according to relevant demographics and the experience of multiple food insecurity episodes.
Methods and Literature Search
This review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis. Searches were performed in December 2020 in PubMed, CINHAL, and Scopus databases and restricted to English language studies published between January 1, 2000, and November 30, 2020. Studies published before 2000 were excluded to maximize homogeneity in food security and child weight status assessment tools. Results were uploaded to the Covidence systematic review tool (Covidence, Melbourne, Australia) for screening.
Studies eligible for inclusion were those with participants who were infants or children in the United States from 1 to 19 years of age; that assessed food security or insufficiency at the household or child level; compared outcomes by food security status; and examined obesity or body mass index (BMI) as the primary outcome of interest (see full search strategy Supplemental Table 3). Included studies were limited to those conducted in the United States to reduce heterogeneity in the food security assessment tools employed. Studies were excluded if they only reported unadjusted relationships between food security and obesity, or if the study population was limited to only youth with overweight or obesity at baseline. Studies with less than 30 participants, the traditional minimum in statistics for a reliable confidence interval, were also excluded. Title and abstract screening, full-text screening and data extraction were performed by 2 independent reviewers; conflicts were resolved by consultation between researchers. Database searches were re-run in September 2021 and February 2022, and results were hand-searched to add relevant studies published between December 2020 and February 2022.
Study quality was assessed using the National Institutes of Health (NIH) quality assessment tool for observational cohort and cross-sectional studies (NIH, Bethesda, Maryland). Although quality assessment tools for clinical trials are well established, there is no consensus on the best methods to assess the quality of observational nutrition studies.9 We selected the NIH tool because it could be applied consistently to all included studies. Two researchers applied the assessment tool independently. Disagreements were resolved via discussion among the research team.
Results
The database search yielded 2272 studies after duplicates were removed. Following title and abstract review, 91 papers were retained for full-text screening. Forty-two were excluded for 1 of the following: methods (no adjusted estimate of the relationship between food security and anthropometrics); outcome (did not assess likelihood of obesity or a continuous BMI-related outcome); population (participants over 19 years old); or location (outside the United States). A total of 41 studies met the inclusion criteria after the initial screening, 4 studies were subsequently added following the search and screening process in September 2021, and no additional studies meeting the inclusion criteria were identified in the February 2022 search (Fig 1).
Preferred Reporting Items for Systematic Reviews and Meta-Analysis flow diagram detailing review search process. PRISMA statement distributed under the terms of the Creative Commons Attribution License. Original source: Mohr D, Liberati A, Tetzlaff J, Altmann DC, the PRISMA Group. Preferred reporting items for systematic reviews and meta-analysis: the PRISMA statement. PLoS Med. 2009;6(7);e1000098.37
Preferred Reporting Items for Systematic Reviews and Meta-Analysis flow diagram detailing review search process. PRISMA statement distributed under the terms of the Creative Commons Attribution License. Original source: Mohr D, Liberati A, Tetzlaff J, Altmann DC, the PRISMA Group. Preferred reporting items for systematic reviews and meta-analysis: the PRISMA statement. PLoS Med. 2009;6(7);e1000098.37
A total of 45 papers were originally included for data extraction. Data on sample size, demographic characteristics of study participants, nutrition assistance program participation (when available), food security assessment methods, outcomes measured, method of analysis, covariates included in analysis, adjusted results, and tests for interaction with any subsequent stratified results, were extracted using a piloted, standardized extraction spreadsheet. Although both longitudinal and cross-sectional studies were initially included, only the longitudinal studies (n = 16) were ultimately retained for evidence analysis because of their overall higher study quality and ability to provide insight into potential relationships between food insecurity and obesity over time.
In the quality analysis, all longitudinal studies were rated “good” or “fair” (Supplemental Table 4). Thirteen studies10–22 used either the “gold standard” USDA 18-item Household Food Security Survey Module or a shorter subset derived directly from the full module to assess food security, whereas 3 studies used single-item measures. These 3 studies were excluded from the overall results synthesis because they lacked a standardized food security assessment instrument.
In the final 13 studies, variation in outcomes measured, food security categorization (eg, binary vs multilevel and categorical versus continuous), and analysis methods prevented us from conducting a meta-analysis. We instead present the results according to the 3 different outcomes analyzed in the included studies. Our primary interest was examining the relationship between food insecurity and obesity. We focused on obesity rather than both overweight and obesity because obesity has a greater sensitivity and specificity for identifying excess body fat and carries a higher risk for adverse health outcomes.23 To examine potential differences in trajectories of BMI growth, we also synthesized the evidence for associations between food insecurity and changes in BMI and BMI z-score. Both variables present interpretation challenges as longitudinal outcomes. BMI changes are not indexed to age and sex-specific references in studies and may not account for the normative dip in BMI that occurs in early childhood.24 BMI z-score changes are smaller at higher levels of adiposity and do not adequately reflect large changes in weight and adiposity.25 Despite these limitations, the growth trajectories identified in these analyses can illuminate associations between food insecurity and excess weight gain in children. When included studies measured multiple outcomes, we included the findings in each of the applicable syntheses. Table 1 summarizes the characteristics of all included studies.
Characteristics of Included Studies
| Source | Number | Location or Cohort | Age Range | Race and Ethnicity | Food Security Assessment Tool | Food Security Level of Assessment | Food Security Categories for Analysis | Covariates Included in Final Model | Outcome(s) Assessed | Main Analysis Findings for Food Insecurity Independent Variable | Tests for Interactions and Findings |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Benjamin-Neelon et al10 (2020) | 666 | Nurture study: birth cohort of women and their infants from 2 clinics in the southeastern United States. | Mothers enrolled during pregnancy, children followed from 1–12 mo | 14.9% White, 68.6% Black, 14.9% non-White, and non-Black or more than 1 race | USDA 18-item HFSSM, but over the past 30 d at each assessment instead of over the past 12 mo | Household; assessed at infant age 3, 6, 9, an d 12 mo | (1) High food security, (2) marginal food security, (3) low food security, and (4) very low food security | Infant race, sex, birth weight for gestational age z-score, breastfeeding, maternal age, race, education, and prepregnancy BMI; number of children in household, Supplemental Nutrition Assistance Program (SNAP) participation, and WIC participation | BMI z-score using WHO reference data | Very low food security associated with increased BMI z-score | Food insecurity x SNAP: null; food insecurity x WIC: null |
| Bhargava et al11 (2008) | 7635 | ECLS-K (national) | Followed from first through fifth grade, mean age in first grade: 73 mo | 9% Black, 16% Hispanic model only controlled for non-Hispanic Black race or Hispanic ethnicity; authors also reported the following demographics: 65% White and 4% Asian | USDA 18-item HFSSM | Household; assessed at first, third, and fifth grades | Continuous variable for aggregate number of affirmative responses (higher score, greater food insecurity) | Child race (binary variables for Black and Hispanic), age (quadratic; not included in BMI z-score model), parental education, number of siblings, health of respondent limited activities (Y/N), emotional wellbeing of respondent, minutes per day TV time, hours per week nonparental care, physical exercise (days per week with > 20 min), and household income lagged dependent variable | BMI, BMI z-score | Null in both models | None reported |
| Bronte-Tinkew et al12 (2007) | 8693 | ECLS-birth cohort (national) | Followed from 9 to 24 mo, age range was 6 to 22 mo for the 9-mo wave; 16 to 38 mo for 24-mo wave | Specific proportions in analysis sample not reported and race not included as covariate | USDA 18-item Household Food Security Survey Module (HFSSM) | Household; assessed at 9-mo wave | (1) High and marginal, (2) low, and (3) very low | Child age, child sex, family structure, parental education, mother’s age at birth, appropriate number of well-baby visits, smoking in household, maternal employment status, household SNAP participation, WIC participation, and household poverty index ratio; mediators: maternal depression, parenting practices, and infant feeding practices | Obesity status at 24 mo (wt-for-length) | Direct relationship: null; mediated relationship: food insecurity associated with poorer positive parenting practices, good positive parenting practices associated with good infant feeding practices, and good infant feeding practices associated with decreased odds of obesity | None reported |
| Burke et al13 (2016) | 15 827 | ECLS-K (national) | Followed from K through eighth grade | Boys, at fall K: 60.5% White, 12.6% Black, 16.7% Hispanic, 10.2% non-White, non-Black, and non-Hispanic; Girls, at fall K: 59.5% White, 13.3% Black, 17.3% Hispanic, 9.9% non-White, non-Black and non-Hispanic | USDA 18-item HFSSM | Household; assessed at fall K, spring K, spring third, spring fifth, and spring eighth | (1) Food secure (high and marginal) and (2) food insecure (low and very low) | Child age, height (quadratic), birth weight, race and ethnicity, health insurance coverage, parental perception of child exercise level (same as, more, or less than other children), school lunch receipt, number of parents in household, urbanicity, region of the United States, SNAP participation, total household size, and socioeconomic status | BMI growth from K to eighth grade using data from fall kindergarten, spring kindergarten, spring third grade, spring fifth grade, and spring eighth grade | Growth curve models stratified by sex: for girls, lower BMI in K and higher BMI in eighth grade for those from food insecure households | Food security status x year: for girls, BMI growth from K to eighth grade was greater for those from food insecure households |
| Gamba et al14 (2021) | 204 | CHAMACOS, Salinas, CA | Mothers enrolled during pregnancy, followed children from age 5 to 12 y for this analysis | 100% Hispanic (mothers) | USDA 6-item Short Form | Household: assessed at child ages 5, 7, 9, and 10.5 | (1) Fully food secure and (2) marginal food security or food insecure, also examined status over multiple time points | Child sex, birth weight (low to normal versus high), early life food insecurity, hours per day TV time, hours per day playing outside, puberty status (prepubescent or pubescent); maternal age (range), marital status, prepregnancy BMI, parity, proportion of life spent in the United States, education level and household income (above or below FPL) | Change in obesity status across age intervals: 5 to 7 y, 7 to 9 y, 9 to 10.5 y, 10.5 to 12 y, change in BMI z-score and BMI over specific age intervals: 5 to 7 y, 7 to 9 y, 9 to 10.5 y, 10.5 to 12 y and change in BMI z-score and BMI across 2 intervals and 3 intervals | Null findings for associations between food insecurity and obesity in all models. Food insecurity at age 9 associated with a decrease in BMI z-score from 9 to 10.5 y, to food security status across 2 intervals, food insecurity across 2 time points associated with decrease in BMI z-score, change from food secure to food insecure across 2 time associated with decreases in both BMI z-score and BMI | None reported in obesity analysis, for BMI z-score and BMI: Food insecurity x maternal prepregnancy BMI: null, food insecurity x child sex: null |
| Jackson et al15 (2017) | 128 | Generating rural options for wt (GROW)-healthy kids and communities: sample from rural communities in 3 Oregon counties | Elementary students followed over 1 y, mean age 8.0 ± 1.9 y at baseline | 89.2% White and 10.8% non-White | 2-item Hunger Vital Signs food security screening tool | Household; assessed at baseline | (1) Not at risk for food insecurity and (2) at risk for food insecurity (affirmative to either item) | Child BMI z-score at baseline, participation in GROW intervention, (main analysis was crosssectional) | Change in BMI z-score from baseline to 1-y follow-up | Null | None reported |
| Jansen et al16 (2017) | 501 | Growing Healthy Study: preschoolers in Head Start programs in urban and rural Michigan | Followed over 6 mo (fall to spring of 1 school year) mean age 4.4 ± 0.6 y at baseline | 60.3% White, 28.7% Black, 6.6% Hispanic, 4.4% non-White, non-Black, and non-Hispanic | USDA 18-item HFSSM | Household; assessed at baseline and follow-up | (1) Food secure (high and marginal) and (2) food insecure (low and very low), created measure using the 2 categories and 2 time points for 4 different food security categories in analysis | Caregiver race and ethnicity, education level, and household income to needs ratio | Change in BMI z-score | Models stratified by child sex: for girls, greater increase in BMI z-score for those whose households became food insecure over the interval | None reported |
| Jyoti et al17 (2005) | 11 460 | ECLS-K (national) | Followed from spring of K to spring of third grade, mean age in spring of K 6.2 y | 62.1% White, 11.4% Black, 16.6% Hispanic, 10.0% non-White, non-Black, and non-Hispanic or more than 1 race | USDA 18-item HFSSM | Household; assessed at K and third grade | (1) Food secure (high) and (2) food insecure (marginal, low, and very low), created measure using the 2 categories and 2 time points for 4 different food security categories in analysis | Child sex, age, birth weight (low or not), race and ethnicity, disability (yes or no), initial height, change in height, BMI in kindergarten, health insurance coverage, exercise frequency (days per week); home language (English or other), household income, number of parents in household, household size, mother’s age, father’s age, parent marital status, mother’s age at first birth, parent employment, parent highest education level, childcare arrangements, number of siblings, parent self-assessment of depression, region of residence, urbanicity, and neighborhood safety rating | Change in BMI from K-3rd grade | When controlling for only kindergarten BMI, food insecurity in kindergarten associated with increased third grade BMI, null when additional covariates are added; food insecurity over time, and food insecurity at both time points associated with increased BMI gain | Food insecurity x sex: increasedBMI gain for girls when experienced food insecurity in K; increased BMI gain for girls if food insecure at both time points |
| Kamdar et al18 (2019) | 137 | Cohort recruited from self-identified Hispanic parent-child dyads enrolled in Head Start in Houston, Texas | Followed children for 18 mo, mean age 4.8 y at baseline and 6.3 y at follow-up | 100% Hispanic (parent) | 5-item, self-administered paper version of USDA 6-item Short Form | Household; assessed at baseline | (1) Food secure (high and marginal) and (2) food insecure (low and very low), interval-level scale score used in analysis | Child BMI z-score at baseline, maternal English acculturation, maternal BMI mediators: parental feeding demandingness, parental feeding responsiveness, and child diet quality at follow-up | BMI z-score | Null (both direct indirect effects) | Food insecurity x maternal English acculturation: null |
| Lee et al19 (2018) | 8167 | ECLS-K:2011 (national | Followed from spring of K to spring of third grade, mean age in fall of K: 5.6 y | 55.9% White, 9.1% Black, 25.6% Hispanic, 0.9% AI and AN, 4.5% Asian or PI, 4.0% Multiracial or unreported | USDA 18-item HFSS | Household; assessed at K and first grad | (1) Food secure (high and marginal) and (2) food insecure (low and very low), separate analyses for status in kindergarten and status in first grade | Sex, race and ethnicity, income quintile, and parental level of education | Odds of obesity in third grade, BMI z-score in third grade | Null, food insecurity in first grade associated with higher BMI z-score in third grade | Food insecurity x WIC participation: null, food insecurity x WIC participation: null |
| Metallinos-Katsaras et al20 (2012) | 28 353 | WIC participants, Massachusetts | Retrospective cohort, followed from first WIC infancy visit to last WIC child visit (2–5 y), mean age at child visit: 36.9 ± 9.7 mo | 41.4% White, 20.6% Black, 31.6% Hispanic, and 6.5% Asian | 4-item subscale of USDA 18-item HFSSM | Household: assessed at infant visit and child visit | (1) Food secure, (2) food insecure without hunger (older term for low food security), and (3) food insecure with hunger (older term for very low food security), created time-integrated measure using 3 categories and 2 time points for 9 different food security categories in analysis | Child sex, race and ethnicity, birth weight, age at the last WIC visit; maternal age, education, prepregnancy wt status, and household size | Odds of obesity at child visit (underweight children excluded), BMI z-score | Food insecurity without hunger at both time points associated with increased and odds of obesity relative to food secure at both time points, no difference in BMI z-score for either food insecure group relative to food security in infancy, food insecurity without hunger in infancy associated with higher BMI z-score at child visit relative to food insecurity with hunger in infancy | Food security status in infancy x maternal prepregnancy wt status: if mothers were either underweight or overweight or with obesity, positive association between food insecurity without hunger at both time points and odds of child obesity relative to food secure at both time points; if mothers were overweight or with obesity, positive association between food insecurity without hunger in infancy and BMI z-score at child visit, relative to food security in infancy |
| Rose and Bodor21 (2006) | 12 890 | ECLS-K (national) | Followed from spring of K to spring of first grade, mean age in spring of K: 74.4 mo | 58.8% White, 15.5% Black, 18.6% Hispanic, 2.7% Asian, 4.4% non-White, non-Black, non-Hispanic, and non- Asian | USDA 18-item HFSSM | Household: assessed at K | (1) Food secure (high and marginal) and (2) food insecure (with and without hunger, ie, low and very low food security) | Child age, sex, race and ethnicity, interaction between sex and race and ethnicity, birth weight, (dummy variables for low or high); household poverty index ratio, maternal education, urbanization, region of the country, frequency of family meals, parent perception of child’s physical activity (relative to other children), and TV time (> 2 h per day or not) | Odds of obesity in first grade, odds of high wt gain between K and first (high wt gain defined as a BMI gain in the top 15% of distribution for age and sex) | Null, food insecurity in K associated with lower odds of high wt gain from K to first grade | None reported for longitudinal analyses |
| Zhu et al22 (2020) | 6368 | ECLS-K (national) | Followed from K-8th grade | Food secure in K: 69.7% White, 7.5% Black, 14.1% Hispanic, 8.6% Asian or non-White, non-Black and non-Hispanic; food insecure in K: 39.1% White, 12.4% Black, 34.1% Hispanic, 14.4% Asian or non-White, non-Black and non-Hispanic, race and ethnicity reported by food security status in kindergarten; significant difference in demographics between food secure and food insecure groups (P < .001) | USDA 18-item HFSSM | Household: assessed at K, third, fifth, and eighth grades | (1) Food secure (high and marginal) and (2) food insecure (low and very low) | Child sex, race and ethnicity, birth weight (low, normal, or high), health insurance status, household poverty (at and above or below FPL), parental depression, maternal nativity (born in U.S. or no), maternal education level | BMI z-score at third, fifth, and eighth grades | Food insecurity in fifth grade associated with higher BMI z-score in eighth grade | Interaction between food security status at adjacent time points: null food security by race and ethnicity: null, food security by household poverty: null, food security by sex: null |
| Only analyses meeting our inclusion criteria are presented in the table; studies may have analyzed other outcomes that did not fit our criteria | |||||||||||
| Source | Number | Location or Cohort | Age Range | Race and Ethnicity | Food Security Assessment Tool | Food Security Level of Assessment | Food Security Categories for Analysis | Covariates Included in Final Model | Outcome(s) Assessed | Main Analysis Findings for Food Insecurity Independent Variable | Tests for Interactions and Findings |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Benjamin-Neelon et al10 (2020) | 666 | Nurture study: birth cohort of women and their infants from 2 clinics in the southeastern United States. | Mothers enrolled during pregnancy, children followed from 1–12 mo | 14.9% White, 68.6% Black, 14.9% non-White, and non-Black or more than 1 race | USDA 18-item HFSSM, but over the past 30 d at each assessment instead of over the past 12 mo | Household; assessed at infant age 3, 6, 9, an d 12 mo | (1) High food security, (2) marginal food security, (3) low food security, and (4) very low food security | Infant race, sex, birth weight for gestational age z-score, breastfeeding, maternal age, race, education, and prepregnancy BMI; number of children in household, Supplemental Nutrition Assistance Program (SNAP) participation, and WIC participation | BMI z-score using WHO reference data | Very low food security associated with increased BMI z-score | Food insecurity x SNAP: null; food insecurity x WIC: null |
| Bhargava et al11 (2008) | 7635 | ECLS-K (national) | Followed from first through fifth grade, mean age in first grade: 73 mo | 9% Black, 16% Hispanic model only controlled for non-Hispanic Black race or Hispanic ethnicity; authors also reported the following demographics: 65% White and 4% Asian | USDA 18-item HFSSM | Household; assessed at first, third, and fifth grades | Continuous variable for aggregate number of affirmative responses (higher score, greater food insecurity) | Child race (binary variables for Black and Hispanic), age (quadratic; not included in BMI z-score model), parental education, number of siblings, health of respondent limited activities (Y/N), emotional wellbeing of respondent, minutes per day TV time, hours per week nonparental care, physical exercise (days per week with > 20 min), and household income lagged dependent variable | BMI, BMI z-score | Null in both models | None reported |
| Bronte-Tinkew et al12 (2007) | 8693 | ECLS-birth cohort (national) | Followed from 9 to 24 mo, age range was 6 to 22 mo for the 9-mo wave; 16 to 38 mo for 24-mo wave | Specific proportions in analysis sample not reported and race not included as covariate | USDA 18-item Household Food Security Survey Module (HFSSM) | Household; assessed at 9-mo wave | (1) High and marginal, (2) low, and (3) very low | Child age, child sex, family structure, parental education, mother’s age at birth, appropriate number of well-baby visits, smoking in household, maternal employment status, household SNAP participation, WIC participation, and household poverty index ratio; mediators: maternal depression, parenting practices, and infant feeding practices | Obesity status at 24 mo (wt-for-length) | Direct relationship: null; mediated relationship: food insecurity associated with poorer positive parenting practices, good positive parenting practices associated with good infant feeding practices, and good infant feeding practices associated with decreased odds of obesity | None reported |
| Burke et al13 (2016) | 15 827 | ECLS-K (national) | Followed from K through eighth grade | Boys, at fall K: 60.5% White, 12.6% Black, 16.7% Hispanic, 10.2% non-White, non-Black, and non-Hispanic; Girls, at fall K: 59.5% White, 13.3% Black, 17.3% Hispanic, 9.9% non-White, non-Black and non-Hispanic | USDA 18-item HFSSM | Household; assessed at fall K, spring K, spring third, spring fifth, and spring eighth | (1) Food secure (high and marginal) and (2) food insecure (low and very low) | Child age, height (quadratic), birth weight, race and ethnicity, health insurance coverage, parental perception of child exercise level (same as, more, or less than other children), school lunch receipt, number of parents in household, urbanicity, region of the United States, SNAP participation, total household size, and socioeconomic status | BMI growth from K to eighth grade using data from fall kindergarten, spring kindergarten, spring third grade, spring fifth grade, and spring eighth grade | Growth curve models stratified by sex: for girls, lower BMI in K and higher BMI in eighth grade for those from food insecure households | Food security status x year: for girls, BMI growth from K to eighth grade was greater for those from food insecure households |
| Gamba et al14 (2021) | 204 | CHAMACOS, Salinas, CA | Mothers enrolled during pregnancy, followed children from age 5 to 12 y for this analysis | 100% Hispanic (mothers) | USDA 6-item Short Form | Household: assessed at child ages 5, 7, 9, and 10.5 | (1) Fully food secure and (2) marginal food security or food insecure, also examined status over multiple time points | Child sex, birth weight (low to normal versus high), early life food insecurity, hours per day TV time, hours per day playing outside, puberty status (prepubescent or pubescent); maternal age (range), marital status, prepregnancy BMI, parity, proportion of life spent in the United States, education level and household income (above or below FPL) | Change in obesity status across age intervals: 5 to 7 y, 7 to 9 y, 9 to 10.5 y, 10.5 to 12 y, change in BMI z-score and BMI over specific age intervals: 5 to 7 y, 7 to 9 y, 9 to 10.5 y, 10.5 to 12 y and change in BMI z-score and BMI across 2 intervals and 3 intervals | Null findings for associations between food insecurity and obesity in all models. Food insecurity at age 9 associated with a decrease in BMI z-score from 9 to 10.5 y, to food security status across 2 intervals, food insecurity across 2 time points associated with decrease in BMI z-score, change from food secure to food insecure across 2 time associated with decreases in both BMI z-score and BMI | None reported in obesity analysis, for BMI z-score and BMI: Food insecurity x maternal prepregnancy BMI: null, food insecurity x child sex: null |
| Jackson et al15 (2017) | 128 | Generating rural options for wt (GROW)-healthy kids and communities: sample from rural communities in 3 Oregon counties | Elementary students followed over 1 y, mean age 8.0 ± 1.9 y at baseline | 89.2% White and 10.8% non-White | 2-item Hunger Vital Signs food security screening tool | Household; assessed at baseline | (1) Not at risk for food insecurity and (2) at risk for food insecurity (affirmative to either item) | Child BMI z-score at baseline, participation in GROW intervention, (main analysis was crosssectional) | Change in BMI z-score from baseline to 1-y follow-up | Null | None reported |
| Jansen et al16 (2017) | 501 | Growing Healthy Study: preschoolers in Head Start programs in urban and rural Michigan | Followed over 6 mo (fall to spring of 1 school year) mean age 4.4 ± 0.6 y at baseline | 60.3% White, 28.7% Black, 6.6% Hispanic, 4.4% non-White, non-Black, and non-Hispanic | USDA 18-item HFSSM | Household; assessed at baseline and follow-up | (1) Food secure (high and marginal) and (2) food insecure (low and very low), created measure using the 2 categories and 2 time points for 4 different food security categories in analysis | Caregiver race and ethnicity, education level, and household income to needs ratio | Change in BMI z-score | Models stratified by child sex: for girls, greater increase in BMI z-score for those whose households became food insecure over the interval | None reported |
| Jyoti et al17 (2005) | 11 460 | ECLS-K (national) | Followed from spring of K to spring of third grade, mean age in spring of K 6.2 y | 62.1% White, 11.4% Black, 16.6% Hispanic, 10.0% non-White, non-Black, and non-Hispanic or more than 1 race | USDA 18-item HFSSM | Household; assessed at K and third grade | (1) Food secure (high) and (2) food insecure (marginal, low, and very low), created measure using the 2 categories and 2 time points for 4 different food security categories in analysis | Child sex, age, birth weight (low or not), race and ethnicity, disability (yes or no), initial height, change in height, BMI in kindergarten, health insurance coverage, exercise frequency (days per week); home language (English or other), household income, number of parents in household, household size, mother’s age, father’s age, parent marital status, mother’s age at first birth, parent employment, parent highest education level, childcare arrangements, number of siblings, parent self-assessment of depression, region of residence, urbanicity, and neighborhood safety rating | Change in BMI from K-3rd grade | When controlling for only kindergarten BMI, food insecurity in kindergarten associated with increased third grade BMI, null when additional covariates are added; food insecurity over time, and food insecurity at both time points associated with increased BMI gain | Food insecurity x sex: increasedBMI gain for girls when experienced food insecurity in K; increased BMI gain for girls if food insecure at both time points |
| Kamdar et al18 (2019) | 137 | Cohort recruited from self-identified Hispanic parent-child dyads enrolled in Head Start in Houston, Texas | Followed children for 18 mo, mean age 4.8 y at baseline and 6.3 y at follow-up | 100% Hispanic (parent) | 5-item, self-administered paper version of USDA 6-item Short Form | Household; assessed at baseline | (1) Food secure (high and marginal) and (2) food insecure (low and very low), interval-level scale score used in analysis | Child BMI z-score at baseline, maternal English acculturation, maternal BMI mediators: parental feeding demandingness, parental feeding responsiveness, and child diet quality at follow-up | BMI z-score | Null (both direct indirect effects) | Food insecurity x maternal English acculturation: null |
| Lee et al19 (2018) | 8167 | ECLS-K:2011 (national | Followed from spring of K to spring of third grade, mean age in fall of K: 5.6 y | 55.9% White, 9.1% Black, 25.6% Hispanic, 0.9% AI and AN, 4.5% Asian or PI, 4.0% Multiracial or unreported | USDA 18-item HFSS | Household; assessed at K and first grad | (1) Food secure (high and marginal) and (2) food insecure (low and very low), separate analyses for status in kindergarten and status in first grade | Sex, race and ethnicity, income quintile, and parental level of education | Odds of obesity in third grade, BMI z-score in third grade | Null, food insecurity in first grade associated with higher BMI z-score in third grade | Food insecurity x WIC participation: null, food insecurity x WIC participation: null |
| Metallinos-Katsaras et al20 (2012) | 28 353 | WIC participants, Massachusetts | Retrospective cohort, followed from first WIC infancy visit to last WIC child visit (2–5 y), mean age at child visit: 36.9 ± 9.7 mo | 41.4% White, 20.6% Black, 31.6% Hispanic, and 6.5% Asian | 4-item subscale of USDA 18-item HFSSM | Household: assessed at infant visit and child visit | (1) Food secure, (2) food insecure without hunger (older term for low food security), and (3) food insecure with hunger (older term for very low food security), created time-integrated measure using 3 categories and 2 time points for 9 different food security categories in analysis | Child sex, race and ethnicity, birth weight, age at the last WIC visit; maternal age, education, prepregnancy wt status, and household size | Odds of obesity at child visit (underweight children excluded), BMI z-score | Food insecurity without hunger at both time points associated with increased and odds of obesity relative to food secure at both time points, no difference in BMI z-score for either food insecure group relative to food security in infancy, food insecurity without hunger in infancy associated with higher BMI z-score at child visit relative to food insecurity with hunger in infancy | Food security status in infancy x maternal prepregnancy wt status: if mothers were either underweight or overweight or with obesity, positive association between food insecurity without hunger at both time points and odds of child obesity relative to food secure at both time points; if mothers were overweight or with obesity, positive association between food insecurity without hunger in infancy and BMI z-score at child visit, relative to food security in infancy |
| Rose and Bodor21 (2006) | 12 890 | ECLS-K (national) | Followed from spring of K to spring of first grade, mean age in spring of K: 74.4 mo | 58.8% White, 15.5% Black, 18.6% Hispanic, 2.7% Asian, 4.4% non-White, non-Black, non-Hispanic, and non- Asian | USDA 18-item HFSSM | Household: assessed at K | (1) Food secure (high and marginal) and (2) food insecure (with and without hunger, ie, low and very low food security) | Child age, sex, race and ethnicity, interaction between sex and race and ethnicity, birth weight, (dummy variables for low or high); household poverty index ratio, maternal education, urbanization, region of the country, frequency of family meals, parent perception of child’s physical activity (relative to other children), and TV time (> 2 h per day or not) | Odds of obesity in first grade, odds of high wt gain between K and first (high wt gain defined as a BMI gain in the top 15% of distribution for age and sex) | Null, food insecurity in K associated with lower odds of high wt gain from K to first grade | None reported for longitudinal analyses |
| Zhu et al22 (2020) | 6368 | ECLS-K (national) | Followed from K-8th grade | Food secure in K: 69.7% White, 7.5% Black, 14.1% Hispanic, 8.6% Asian or non-White, non-Black and non-Hispanic; food insecure in K: 39.1% White, 12.4% Black, 34.1% Hispanic, 14.4% Asian or non-White, non-Black and non-Hispanic, race and ethnicity reported by food security status in kindergarten; significant difference in demographics between food secure and food insecure groups (P < .001) | USDA 18-item HFSSM | Household: assessed at K, third, fifth, and eighth grades | (1) Food secure (high and marginal) and (2) food insecure (low and very low) | Child sex, race and ethnicity, birth weight (low, normal, or high), health insurance status, household poverty (at and above or below FPL), parental depression, maternal nativity (born in U.S. or no), maternal education level | BMI z-score at third, fifth, and eighth grades | Food insecurity in fifth grade associated with higher BMI z-score in eighth grade | Interaction between food security status at adjacent time points: null food security by race and ethnicity: null, food security by household poverty: null, food security by sex: null |
| Only analyses meeting our inclusion criteria are presented in the table; studies may have analyzed other outcomes that did not fit our criteria | |||||||||||
Only analyses meeting our inclusion criteria are presented in the table; studies may have analyzed other outcomes that did not fit our criteria
Food Insecurity and Obesity
Five studies examined the association between food insecurity and obesity.12,14,19–21 Among a cohort of participants in the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC) in Massachusetts, an association was found between low food security (worried about food but no disruption to eating patterns, as opposed to very low food security, which also includes disruption of eating patterns) when present in both infancy and preschool and greater odds of obesity in preschool.20 This relationship was also moderated by maternal prepregnancy weight status, with greater odds of obesity for children in households with low food security at both time points and whose mothers were either overweight, had obesity, or were underweight prepregnancy. An analysis of the Early Childhood Longitudinal Study (ECLS) birth cohort using structural equation modeling found no direct association between food insecurity and obesity, but in mediation analysis, food insecurity affected parenting and infant feeding behaviors, which ultimately affected weight.12 The 3 remaining studies found no significant associations between food insecurity and obesity; 2 used data from the ECLS-K21 and ECLS-K: 201119 cohorts, and the third used findings from a cohort of Hispanic mothers and their children in California (CHAMACOS).14
Food Insecurity and BMI
Four studies analyzed the relationship between food insecurity and changes in BMI over time. Of the 2 studies using ECLS-K data, 1 found a greater increase in BMI among children whose households were food insecure at 2 time points,17 and both found higher BMI increases for girls from food insecure households but not for boys.13,17 In a third study using ECLS-K data, the authors reported no significant associations but did not test for interactions by sex or consider food insecurity at multiple time points.11 The CHAMACOS study found lower BMI gains among children whose households changed from highly food secure to marginally food secure or food insecure across 2 time points.14
Food Insecurity and BMI z-Score
Nine studies investigated associations between food insecurity and changes in BMI z-score. The ECLS-K: 2011 study found an association between food insecurity in first grade and an increased BMI z-score in third grade, but no association between kindergarten food insecurity and third grade BMI z-score.19 Likewise, a fourth ECLS-K study found an association between fifth grade food insecurity and a higher eighth grade BMI z-score but no significant associations when food insecurity occurred in younger grades.22 In a Head Start preschool cohort, an association was found between food insecurity and increased BMI z-score for girls, but participants were only followed for an average of 6 months.16 A birth cohort following infants through 12 months found an association between very low food security and a higher BMI z-score,10 whereas the WIC cohort found no association in the main analysis but an association between food insecurity in infancy and an increased BMI z-score in early childhood if the mother was overweight or had obesity prepregnancy.20 In the CHAMACOS cohort, food insecurity at age 9 was associated with a decrease in BMI z-score from ages 9 to 10.5, and food insecurity across 2 time points or changing from food secure to food insecure were also associated with decreased BMI z-score.14 The remaining 3 studies found no significant associations.11,15,18
Results Synthesis
The studies are categorized by outcome and findings in Table 2. An association between food insecurity and obesity was found only in early childhood,20 whereas 6 additional studies found evidence of associations between food insecurity and increases in BMI or BMI z-score in limited age groups or sex-specific analyses.10,13,16,17,19,22 One study of an exclusively Hispanic population found evidence of an association between food insecurity and decreases in BMI z-score or BMI, limited to a specific age group or changes in food security status.14 Although all studies assessed food security based on standardized US Department of Agriculture assessment tools, comparison is challenging because of differences in food security categorization. Most studies categorized participants as either food secure (high or marginal food security according to survey responses) or food insecure (low or very low food security). However, 2 studies combined marginal food security with low and very low food security,14,17 3 studies used more than 2 categories for food security status,10,12,20 and 1 used a continuous variable.11 Studies also differed in the covariates used in their analyses. Child age, sex, race or ethnicity, household income, and parent or maternal education were consistently included as control variables, but other predictors of obesity, such as physical activity level, child birth weight, and maternal BMI, were included in no more than half of the studies. The variations in both the food security variable and covariates may help explain the mixed results observed across studies.
Findings by Outcome
| Outcome | Number of Studiesa | Increase | Decrease | No Association |
|---|---|---|---|---|
| Odds of obesity | 5 | Metallinos-Katsaras et al20 (2012) (WIC) | Bronte-Tinkew et al12 (2007) (ECLS-B), Gamba et al14 (2021) (CHAMACOS), Lee et al19 (2018) (ECLS-K:2011), Rose and Bodor21 (2006) (ECLS-K) | |
| Changes in BMI | 4 | Burke et al13 (2016) (ECLS -K), Jyoti et al17 (2005) (ECLS-K) | Gamba et al14 (2021) (CHAMACOS) | Bhargava et ak11 (2008) (ECLS-K) |
| Changes in BMI z-score | 9 | Benjamin-Neelon et al10 (2020) (birth cohort), Jansen et al16 (2017) (Head Start), Lee et al19 (2018) (ECLS-K:2011), Metallinos-Katsaras et al20 (2012) (WIC), Zhu et al22 (2020) (ECLS-K) | Gamba et al14 (2021) (CHAMACOS) | Bhargava et al11 (2008) (ECLS-K), Jackson et al15 (2017) (elementary, rural), Kamdar et al18 (2019) (Head Start, Hispanic) |
| Outcome | Number of Studiesa | Increase | Decrease | No Association |
|---|---|---|---|---|
| Odds of obesity | 5 | Metallinos-Katsaras et al20 (2012) (WIC) | Bronte-Tinkew et al12 (2007) (ECLS-B), Gamba et al14 (2021) (CHAMACOS), Lee et al19 (2018) (ECLS-K:2011), Rose and Bodor21 (2006) (ECLS-K) | |
| Changes in BMI | 4 | Burke et al13 (2016) (ECLS -K), Jyoti et al17 (2005) (ECLS-K) | Gamba et al14 (2021) (CHAMACOS) | Bhargava et ak11 (2008) (ECLS-K) |
| Changes in BMI z-score | 9 | Benjamin-Neelon et al10 (2020) (birth cohort), Jansen et al16 (2017) (Head Start), Lee et al19 (2018) (ECLS-K:2011), Metallinos-Katsaras et al20 (2012) (WIC), Zhu et al22 (2020) (ECLS-K) | Gamba et al14 (2021) (CHAMACOS) | Bhargava et al11 (2008) (ECLS-K), Jackson et al15 (2017) (elementary, rural), Kamdar et al18 (2019) (Head Start, Hispanic) |
The number of studies does not sum to 13 because some studies investigated multiple outcomes.
Discussion
Our findings corroborate the previously published literature, indicating that potential relationships between food insecurity and childhood obesity and child weight changes are complex. Although the evidence did not allow us to draw broad conclusions about the relationship between food insecurity and obesity in children, we were nevertheless able to gain deeper insight and identify directions for further research by synthesizing results according to age, sex, and multiple experiences of food insecurity.
We observed associations between food insecurity and increases in BMI or BMI z-score among infants,10 preschoolers,16,20 elementary students,17,19 and middle school students.13,22 Although 5 of the studies with evidence of higher BMIs among food insecure youth were highly powered cohorts with large samples,13,17,19,20,22 findings were limited to specific age ranges or subgroups within the sample, with the exception of the preschool study.20 In the localized CHAMACOS cohort, food insecurity was associated with decreased BMI z-scores during mid- to late elementary years.14 Thus, the mixed evidence is in agreement with the 2015 Dietary Guidelines Advisory Committee (DGAC) conclusion that limited evidence supports an association between food insecurity and higher anthropometric measurements in early childhood.26 It is also particularly noteworthy that none of the studies followed children beyond eighth grade. Eight of the 13 studies in our review were published after the 2015 DGAC identified the need for additional study of food insecurity and weight changes into the adolescent years,26 but none provided evidence of potential associations beyond middle school.
Three of the studies in the review presented results stratified by child sex, and all found an association between food insecurity and increased BMI or BMI z-score for girls but not for boys.13,16,17 Three additional studies tested for interaction by sex but found no associations.14,20,22 Associations between food insecurity and higher BMI for preschool girls have also been found cross-sectionally.27,28 Potential explanations for this association in girls but not in boys could include differential parent feeding practices by gender,29 or different responses to stress, including the experience of food insecurity.17 Lack of testing for interaction by sex in many studies could also be masking associations in 1 of the groups, even if a relationship is not found in the overall population.27 Following youth into adolescence and early adulthood could also help clarify differences in the interactions between food insecurity and weight by sex, particularly given recent evidence among adults that food insecurity was more prevalent among women with greater adiposity.30
Several of the longitudinal studies in the review categorized food security across multiple time points to examine how changes in food security status or multiple episodes of food insecurity were related to obesity and BMI.13,14,16,17,20 In 3 large studies, food insecurity at multiple time points was associated with obesity or greater BMI growth relative to food security at all time points,13,17,20 and 1 preschool study found that for girls, higher BMI z-scores were associated with the household changing from food secure to food insecure over the course of 1 school year.16 One smaller study found an association with decreased BMI z-scores when food insecurity occurred at multiple time points or when households transitioned from food security to food insecurity,14 but more evidence points to a potential cumulative positive effect of multiple experiences of food insecurity on weight gain, an effect also observed by the 2015 DGAC.26 The effects of the duration and episodic nature of food insecurity may be of particular relevance to the increase in childhood obesity that has occurred during the COVID-19 pandemic.6,7
Potential differences by age range, sex, and the unknown effects of fluctuations in food security status over time indicate that a systems or structural modeling approach may provide better insight into how food insecurity and child weight status are related to one another through indirect channels. Household stress may play an important mediating role in the relationship between food insecurity and weight outcomes. One longitudinal study with a small sample size found an association between food insecurity and increased BMI when high stress was present at the child level,31 and multiple studies have examined how interactions between maternal stress and food insecurity may affect child weight status.32–35 Two of the studies included in this review included structural models that accounted for parental feeding practices,12,18 and another structural model includes child dietary intake and both child and parent physical activity levels.36 Further research can build on such models to better understand the complex mechanisms that affect the relationship between food insecurity and child weight status. Irrespective of any future conclusive evidence on the relationship mechanisms between food insecurity and childhood obesity, however, effective interventions against child food insecurity should be a public health priority to promote the physical, emotional, and cognitive wellbeing of children and parents.
Limitations
Although limiting our analysis to longitudinal studies strengthens the evidence relative to cross-sectional findings, following low-income populations over long periods is a challenging endeavor. The ECLS-K studies did not remain nationally representative over the follow-up periods, and 1 of them specifically noted that participants excluded because of missing data were more likely to be of lower socioeconomic status.19 Recent evidence indicates that racial or ethnic disparities in childhood obesity have increased since the COVID-19 pandemic,6 but our ability to explore potential differences by race or ethnicity in the food insecurity-childhood obesity relationship was limited by lack of testing for interaction by race or ethnicity. Two of the 3 studies that followed youth into puberty omitted any discussion of pubertal status, despite connections between puberty and anthropometric measurements that could have affected study findings. Differences in covariates, most notably omission of control variables for physical activity in most studies and for dietary quality in all studies, may contribute to the inconsistent findings. Finally, we were limited in our ability to assess the relationship between food insecurity and obesity by the diverse outcomes measured in the included studies. A greater proportion of studies used continuous BMI outcomes relative to weight categories. Although these studies showed changes in BMI trajectories, it was not apparent whether these changes indicated movement across weight categories.
Conclusions
We observed mixed evidence of associations between food insecurity and childhood obesity, but the mechanisms of their relationship remain difficult to ascertain. This review highlights the importance of understanding the many nuances of how food insecurity and childhood obesity interact with one another, which is even more critical as we have observed increased child food insecurity and widening disparities in the prevalence of obesity amid the COVID-19 pandemic. Ongoing and future studies need to consider interactions between food insecurity and salient demographics and the broader context of the household environment to enable us to meet the dual challenges of reducing childhood obesity and ensuring food security for all families.
Acknowledgments
We thank the anonymous reviewers for their thoughtful and insightful feedback on this paper.
Ms St. Pierre and Dr Ver Ploeg conceptualized and designed the study, coordinated and supervised data collection, drafted the initial manuscript, and reviewed and revised the manuscript; Drs Dietz and Sacheck conceptualized and designed the study and critically reviewed the manuscript for important intellectual content; Ms Pryor, Ms Jakazi, Ms Layman, Ms Noymer, and Ms Coughtrey-Davenport collected data, conducted the initial analyses, and reviewed and revised the manuscript; and all authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.
FUNDING: This research was supported by Healthy Eating Research, a national program of the Robert Wood Johnson Foundation.
CONFLICT OF INTEREST DISCLOSURES: The authors have indicated they have no conflicts of interest to disclose.
Comments