![[Feedback]](/icons/banner/feedbackACT.gif){kind=icon}
![[For Subscribers]](/icons/banner/subscriberACT.gif){kind=icon}
![[Archive]](/icons/banner/archiveACT.gif){kind=icon}
![[Search]](/icons/banner/searchACT.gif){kind=icon}
![[Contents]](/icons/banner/tocACT.gif){kind=icon}
|
|
|
|
|
||
Mathematica Policy Research, Inc., Princeton, NJ 08543-2393
* To whom correspondence should be addressed. E-mail: rwhitaker{at}temple.edu.
 |
ABSTRACT |
|---|
 TOP ABSTRACT IntroductionMethodsResultsDiscussionLITERATURE CITED |
|---|
30 kg/m2) and 31% were not food secure. After adjusting for sociodemographic characteristics and baseline BMI, there were no significant differences in 2-y weight increases between 4 groups that differed in food security status: food secure at both time points (n = 1000), 1.7 kg (95% CI = 1.1–2.3); food secure at baseline, but not at follow-up (n = 183), 2.1 kg (95% CI = 0.7–3.5); not food secure at either time point (n = 257), 1.7 kg (95% CI = 0.5–2.9); and not food secure at baseline but food secure at follow-up (n = 267), 1.9 kg (95% CI = 0.7–3.0). In this population of urban women, changes in food security status over 2 y were not significantly associated with changes in weight. These findings do not support a causal association between food insecurity and obesity.
|
Introduction |
|---|
|
TOPABSTRACTIntroductionMethodsResultsDiscussionLITERATURE CITED |
|---|
Several mechanisms have been advanced to explain how food insecurity might lead to obesity (7). For example, food insecurity could cause individuals to increase their consumption of inexpensive foods that are high in energy density (8) or to overeat in the face of an unpredictable food supply (9,10). Despite the plausibility of these mechanisms, there are no data establishing a causal association between food insecurity and obesity. Most studies of the association between food insecurity and adult obesity have been cross sectional. To our knowledge, the only longitudinal study examined changes in self-reported body weight rather than measured body weight (11). Short of an experimental or quasi-experimental study that alters food security status in an effectively random fashion, the strongest research design to determine whether food insecurity causes excess weight gain is a longitudinal study that measures both food security status and weight at 2 or more time points and examines the association between changes in food security status and changes in weight.
Using data from a longitudinal cohort study of women living in 20 large U.S. cities, we analyzed the relationship between changes in food security and changes in weight over a 2-y period. Our primary hypothesis was that among women who were food secure at baseline, those who were no longer food secure at follow-up would gain more weight than those who remained food secure. Our secondary hypothesis was that among those who were not food secure at baseline, those who became food secure would gain less weight than those who were still not food secure at follow-up.
|
Methods |
|---|
|
TOPABSTRACTIntroductionMethodsResultsDiscussionLITERATURE CITED |
|---|
Women were surveyed in person shortly after delivery and were surveyed by phone 
1, 3, and 5 y later. At 3 and 5 y, an in-home assessment was also attempted for the entire cohort. Of the 4898 women in the original cohort, 1751 agreed to participate in the in-home assessments at both 3 and 5 y and had their height and weight measured. This report uses data collected on these women at 3 and 5 y to examine the relationship between changes in food security status and changes in weight over this 2-y period. Hereafter, we refer to the data from y 3 and 5 as "baseline" and "follow-up," respectively.
Height and weight. Subjects were measured while wearing light clothing and no shoes. Weight was obtained with an electronic scale (SECA 840 Bella Digital Scale) and height was obtained with a portable stadiometer (SECA 214 Road Rod Stadiometer). All subjects were weighed unless they were pregnant, exceeded the scale limit of 140 kg, or were unwilling to be weighed. In each of these situations, the woman was asked to report her current weight or to report her prepregnant weight if she was pregnant. At baseline and follow-up, 122 (7.0%) and 97 (5.5%) women, respectively, were pregnant and an additional 164 (9.4%) and 104 (5.9%) women, respectively, reported their weight rather than being measured. In the baseline assessment in 2 of the 20 cities, heights were reported rather than measured, but in all other assessments, heights were measured. Those who were unwilling to have their height measured at either assessment were asked to report their height.
From these measurements, we calculated BMI (kg/m2), and those with a BMI 30 were classified as obese. The height used in each BMI calculation was the mean of the available heights from the 2 assessments, but the measured height was used if it was only available from 1 assessment. For 46 (2.6%) of the 1751 women, the reported height rather than the measured height was used to calculate BMI.
Food security. Subjects completed the U.S. Household Food Security Survey Module at baseline and follow-up (13). This instrument was developed, refined, and validated during more than a decade of research (14). Our measure of food security was based on the 10 household- and adult-referenced questions in the module (Supplemental Table 1). We did not include the 8 child-referenced questions in the module, because our study was focused on the mother's experience of household food insecurity and how it might affect her body weight.
The 10 items ask about the prior 12 mo and assess food security along a continuum by asking about a range of conditions and behaviors experienced when the quality or amount of available food is constrained due to lack of money or other resources. We computed the number of positive responses, the total number of food insecure conditions reported, using previously described procedures to score item responses as positive and to code any missing responses (13). The term "fully food secure" is customarily used to refer to those with no positive items. For ease of presentation, we refer to these individuals as "food secure" and refer to those with any positive items as "not food secure." Following other published studies (15–17), in some analyses, we further classify those who were not food secure into 2 groups: "marginally food secure" (1 or 2 positive responses) and "food insecure" (3 or more positive responses). This 3-level classification was used to facilitate comparisons between our data and those previously published by others.
Covariates. From the baseline data, we selected 7 sociodemographic covariates that were potentially associated with both the key exposure (food security) and the key outcome (weight) and that might, therefore, act as confounders in our analysis. The income:poverty ratio was calculated as the household income divided by the income at the federal poverty threshold for the year of the income report and the household size. In addition, we used data on race/ethnicity, education level, relationship/marital status, number of children, and months of employment in the prior year. Finally, women were considered to have experienced material hardship (other than food insecurity) during the preceding 12 mo if, due to lack of money, they had been evicted from their home or apartment, had service turned off by the gas or electric company, or had been refused the delivery of heating oil.
Statistical analysis. The final analytic sample contained 1707 subjects. Of the 1751 subjects who had BMI measured at baseline and follow-up, we excluded 36 who were missing data on food security status at 1 or both time points and an additional 8 who were missing data on 1 or more of the 7 covariates. Using sociodemographic data from the survey conducted at the time of delivery, we compared the characteristics of the 1707 women who are the focus of this study to the other 3191 women in the original cohort. Chi-square tests and t tests were used for these comparisons.
In our analysis, we first described the BMI levels and food security status of our sample and, using chi-square tests, examined the relationship of these variables at baseline to our sociodemographic covariates. To make our analyses comparable to those conducted in other studies, we used chi-square tests to examine the cross-sectional association between food security status and obesity at baseline as well as the longitudinal association of food security status at baseline and obesity at follow-up. To adjust these associations for our sociodemographic covariates, we used logistic regression models with obesity as the dependent variable.
We examined the outcome of weight change using 3 categorical variables. One outcome variable consisted of 4 categories of weight change: loss (–2.0 kg), stable (–1.9 to 1.9 kg), moderate gain (2.0–4.9 kg), and large gain (5.0 kg). In addition, we examined 2 binary outcome variables, weight gain of 5 kg and weight gain of 2 kg. Prior research has shown that weight gain at these levels is associated with clinically meaningful changes in metabolic parameters, such as blood pressure or blood glucose, which are related to obesity (18–20). Using chi-square tests, we examined changes in weight over 2 y by baseline food security status and BMI.
To assess change in food security status between baseline and follow-up, subjects were placed into 1 of the following 4 groups: 1) food secure at both times; 2) food secure at baseline but not at follow-up; 3) not food secure at either time; and 4) not food secure at baseline but food secure at follow-up. Our hypotheses involved 2 types of comparisons: between groups 1 and 2 (hypothesizing greater weight gain in group 2) and between groups 3 and 4 (hypothesizing less weight gain in group 4).
We used chi-square tests to compare the distribution of weight change between those who changed food security status and those who did not (between subjects in groups 1 and 2 and between subjects in groups 3 and 4). Then, using logistic regressions models with weight gain 5 kg as the dependent variable, we compared the risk of this outcome for those who changed food security status to those who did not, while adjusting for baseline BMI and sociodemographic covariates. We repeated these models using weight gain 2 kg as the dependent variable and again excluding cases for which height and weight were reported rather than measured. Finally, we used linear regression models to compare the mean weight changes between the 4 groups, while adjusting for covariates.
Statistical analyses were conducted using SPSS v 15 and STATA v 8. P-values smaller than 0.05 were considered significant and the sample means were presented with SD.
|
Results |
|---|
|
TOPABSTRACTIntroductionMethodsResultsDiscussionLITERATURE CITED |
|---|
The prevalence of obesity was 41 and 45% at baseline and follow-up, respectively, whereas 31 and 26% of women had at least 1 positive response to the 10 food security questions (Table 1). The mean weight change in the 2 y between baseline and follow-up was 1.8 kg ± 9.9 kg, with 47% gaining 2 kg and 29% gaining 5 kg. Of the 1183 women who were food secure at baseline, 183 (15.5%) were no longer food secure at follow-up, and of 524 women who were not food secure at baseline, 267 (51%) were food secure at follow-up.
|
|
|
2 kg was 46, 52, and 47% (P = 0.16), respectively, and the percentage of women gaining 5 kg was 28, 32, and 28% (P = 0.35), respectively. Weight gains of these 3 groups were: 1.8 ± 9.9 kg, 1.8 ± 9.6 kg, and 1.6 ± 10.5 kg (P = 0.97), respectively. Weight change was significantly related to baseline BMI (Table 4), with the obese gaining less weight than the nonobese (0.1 ± 11.9 vs. 2.9 ± 8.1 kg; P < 0.001).
|
5 kg were not significantly different for those who changed food security status and those who did not (Table 6). In logistic regression models, these OR were not meaningfully changed with the addition of baseline BMI or the sociodemographic covariates (Table 6). These logistic regression models still showed no significant association between change in food security status and change in weight when we (1) used an outcome of gaining 2 kg or (2) excluded all cases for which height and weight were reported rather than measured (data not shown).
|
|
|
Discussion |
|---|
|
TOPABSTRACTIntroductionMethodsResultsDiscussionLITERATURE CITED |
|---|
5 kg than those who remained food secure. Although baseline food security status was associated with obesity at baseline and obesity at follow-up, these associations were no longer significant after controlling for sociodemographic factors. Comparison with prior studies. Several cross-sectional studies have found that obesity is more common in women who report food insecurity or insufficiency than in those who do not (2–6). However, some studies have not found this association in women (21,22) and some that have found it in women did not find it in men (4,6). We did find an association between baseline food security status and obesity that was of similar magnitude to that found by others (2,4,5), suggesting at least some similarities between our study population and those in other studies. As in the study of Laraia et al. (21), however, our cross-sectional findings were weaker and no longer significant after adjustment for a number of sociodemographic covariates.
To our knowledge, this is only the second longitudinal study in women to examine the relationship between changes in both food security status and body weight. Using nationally representative data on 5303 women 18–74 y of age, Jones and Frongillo (11) examined changes in food security status and changes in self-reported weight over a 2-y period (1999–2001). Like us, they found no difference in weight gain between those women who remained food secure and those who were food secure at baseline but not at follow-up. In a separate report using the same data, they also showed that food security status at baseline was not associated with any increased risk of gaining >2.3 kg (23).
Using data from the NHANES (1999–2002), Wilde and Peterman (6) found that women (but not men) who reported being marginally food secure during the prior 12 mo had an increased risk of gaining 4.54 kg (10 pounds) over that time period compared with those who reported being fully food secure (OR 1.68, 95% CI = 1.21–1.23). The different study populations may explain why our results differ. In addition, the data in their study was obtained for each subject at 1 assessment and the authors computed weight gain as the difference between the subjects' current self-reported weight and the weight they recalled being 12 mo earlier.
Limitations. Although our sample was drawn from 15 states, it was limited to women living in large cities and, by design, many were unwed mothers. Thus, our findings are not meant to apply to all U.S. women. The study involved only about one-third of the mothers from the original birth sample. Those we studied were slightly more socioeconomically disadvantaged than the remainder of the cohort and we cannot determine how these differences might have influenced our findings about the relationship between food security and weight change. We cannot exclude the possibility of some selection bias.
Although we did not find significant associations between weight gain over 2 y and either baseline food security status or change in food security status, our sample may have been too small to detect group differences that might be regarded by some as meaningful from a policy or clinical perspective. In addition, our sample was too small to determine whether the relationship between changes in food security status and changes in weight differed by certain characteristics, such as income or baseline obesity status. Finally, height and/or weight were self-reported in some subjects, but our results were unchanged when we excluded these subjects from our analysis.
Interpretation and implications. Our findings are not consistent with the idea that food insecurity is a cause of excessive weight gain and obesity and the findings were the same with different analytic approaches, such as using continuous or categorical measures of weight change and excluding cases with self-reported weight. However, there are a few potential reasons that we could have missed a true association between changes in food security and changes in weight. Although food security status and obesity can be assessed in individuals at any given time point, individuals develop and experience these conditions over time and their association cannot necessarily be elucidated in a study in which each is measured at only 2 points in time. Obesity is a chronic condition, which develops over a period of months to years and food insecurity can be episodic or chronic. There may be a long lag in the effects of food security on obesity.
Although we used the standard measure of food security, it has known limitations (24). In our study, women reported on food security during a 12-mo period before each BMI assessment and the possibility exists that important variation in food security status between and within individuals was not captured by this method. It is possible, for example, that food insecurity leads to weight change only when it is experienced for a long duration and that some of the women in our study that had a reduction in food security only experienced this reduction for 1 y or less before their follow-up BMI measurement.
Finally, it is possible that food insecurity promotes excess weight gain only in some individuals. For example, the factors favoring this association in women more than men are not known. Among women, there may be factors, such as the disposition toward binge eating, which place some women at greater risk for gaining weight in the setting of food insecurity.
In conclusion, our longitudinal study of urban women does not suggest that food insecurity causes excess weight gain. Future studies examining the relationship between food insecurity and obesity should be longitudinal, with multiple measures of both food security status and weight, or they should be experimental or quasi-experimental in design.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
2 Author disclosures: R. C. Whitaker and A. Sarin, no conflicts of interest.
3 Present address: Center for Obesity Research and Education, Temple University, Philadelphia, PA 19140.
Manuscript received 23 April 2007. Initial review completed 18 May 2007. Revision accepted 17 June 2007.
|
LITERATURE CITED |
|---|
|
TOPABSTRACTIntroductionMethodsResultsDiscussionLITERATURE CITED |
|---|
1. Life Sciences Research Office, Federation of American Societies for Experimental Biology. Core indicators of nutritional state for difficult-to-sample populations. J Nutr. 1990;120 Suppl 11:1559–600.[Medline]
2. Adams EJ, Grummer-Strawn L, Chavez G. Food insecurity is associated with increased risk of obesity in California women. J Nutr. 2003;133:1070–4.
3. Olson CM. Nutrition and health outcomes associated with food insecurity and hunger. J Nutr. 1999;129:S521–4.[Medline]
4. Townsend MS, Peerson J, Love B, Achterberg C, Murphy SP. Food insecurity is positively related to overweight in women. J Nutr. 2001;131:1738–45.
5. VanEenwyk J, Sabel J. Self-reported concern about food security associated with obesity: Washington, 1995–1999. MMWR Morb Mortal Wkly Rep. 2003;52:840–2.[Medline]
6. Wilde PE, Peterman JN. Individual weight change is associated with household food security status. J Nutr. 2006;136:1395–400.
7. Olson CM. Food insecurity in women: a recipe for unhealthy trade-offs. Top Clin Nutr. 2005;20:321–8.
8. Drewnowski A, Specter SE. Poverty and obesity: the role of energy density and energy costs. Am J Clin Nutr. 2004;79:6–16.
9. Polivy J. Psychological consequences of food restriction. J Am Diet Assoc. 1996;96:589–92.[Medline]
10. Kendall A, Olson CM, Frongillo EA Jr. Relationship of hunger and food insecurity to food availability and consumption. J Am Diet Assoc. 1996;96:1019–24.[Medline]
11. Jones SJ, Frongillo EA. The modifying effects of Food Stamp Program participation on the relation between food insecurity and weight change in women. J Nutr. 2006;136:1091–4.
12. Reichman NE, Teitler JO, Garfinkel I, McLanahan SS. Fragile families: sample and design. Child Youth Serv Rev. 2001;23:303–26.
13. Bickel G, Nord M, Price C, Hamilton W, Cook J. Guide to measuring household food security, revised 2000. Alexandria (VA): USDA, Food and Nutrition Service; 2000.
14. Carlson SJ, Andrews MS, Bickel GW. Measuring food insecurity and hunger in the United States: development of a national benchmark measure and prevalence estimates. J Nutr. 1999;129:S510–6.[Medline]
15. Nord M, Andrews M, Carlson S. Household food security in the United States, 2005. Alexandria (VA): Economic Research Service, USDA; 2005. Economic Research Report No.: 29.
16. Winicki J, Jemison K. Food insecurity and hunger in the kindergarten classroom: its effect on learning and growth. Contemp Econ Policy. 2003;21:145–57.
17. Jyoti DF, Frongillo EA, Jones SJ. Food insecurity affects school children's academic performance, weight gain, and social skills. J Nutr. 2005;135:2831–9.
18. Haffner SM. Relationship of metabolic risk factors and development of cardiovascular disease and diabetes. Obesity (Silver Spring). 2006;14 Suppl 3:S121–7.[Medline]
19. Hamman RF, Wing RR, Edelstein SL, Lachin JM, Bray GA, Delahanty L, Hoskin M, Kriska AM, Mayer-Davis EJ, et al. Effect of weight loss with lifestyle intervention on risk of diabetes. Diabetes Care. 2006;29:2102–7.
20. Willett WC, Dietz WH, Colditz GA. Guidelines for healthy weight. N Engl J Med. 1999;341:427–34.
21. Laraia BA, Siega-Riz AM, Evenson KR. Self-reported overweight and obesity are not associated with concern about enough food among adults in New York and Louisiana. Prev Med. 2004;38:175–81.[Medline]
22. Basiotis PP, Lino M. Food insufficiency and prevalence of overweight among adult women. Fam Econ Nutr Rev. 2003;15:55–7.
23. Jones SJ, Frongillo EA. Food insecurity and subsequent weight gain in women. Public Health Nutr. 2007;10:145–51.
24. National Research Council, Committee on National Statistics, Division of Behavioral and Social Sciences and Education. Food insecurity and hunger in the United States: an assessment of the measure. Washington: National Academies Press; 2006.
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
