Abstract
Background: The Supplemental Nutrition Assistance Program (SNAP) is the largest federal nutrition assistance program. In recent years, SNAP participation rates increased during times of economic hardship.
Objective: We examined whether household SNAP participation was associated with adiposity and metabolic risk factors in a representative sample of low-income US adults.
Design: A cross-sectional analysis was performed with the use of data from the 2003–2006 National Health and Nutrition Examination Surveys. The study population was restricted to nonelderly adults whose household incomes fell to or <130% of the federal poverty level. Multinomial logistic and Poisson regression models were fit to examine the associations between SNAP participation and BMI, waist circumference, and metabolic risk factors among 2250 low-income adults.
Results: In the previous 12 mo, 32.8% of adults received household SNAP benefits. SNAP participation was positively associated with obesity [prevalence ratio (PR): 1.58; 95% CI: 1.08, 2.31], waist circumference in men (PR for top compared with bottom quartile: 2.04; 95% CI: 1.15, 3.62; P = 0.02), and waist circumference in women (PR: 2.95; 95% CI: 1.51, 5.77; P = 0.003; P-interaction with sex = 0.11), independent of sociodemographic characteristics. SNAP participation was also related to elevated triglycerides (PR: 1.71; 95% CI: 1.33, 2.20), lower HDL cholesterol (PR: 1.23; 95% CI: 1.08, 1.41), elevated fasting glucose (≥110 mg/dL; PR: 1.63; 95% CI: 1.05, 2.52), and metabolic syndrome (PR: 1.49; 95% CI: 1.13, 1.95). Associations with triglycerides and HDL cholesterol persisted after adjustment for BMI.
Conclusion: Household SNAP participation was positively associated with BMI, waist circumference, and metabolic risk factors among low-income adults. These associations may be mediated by dietary intake and warrant further investigation.
INTRODUCTION
The federal SNAP4, formerly the FSP, aims to alleviate food insecurity and improve the dietary intake of participants through benefits to purchase household food items. SNAP eligibility is determined by having a household income at or below 130% of the FPL and <$2000 in countable assets (1). In recent years, participation in SNAP has increased dramatically because more households have fallen into poverty (2). In 2010, $68 billion was spent on SNAP, with 40.3 million persons receiving benefits (2). Between July 2010 and July 2011, there was a 10.4% increase in SNAP participation among US households (3).
A 2007 USDA report reviewed several previous studies exploring the relation between SNAP participation and obesity (4). Among children, there appeared to be no association between FSP participation and overweight or obesity; these findings were relatively consistent across data sets and analytic methods. Among adults, longitudinal studies using fixed-effect (5, 6), discrete factor (7), and bivariate probit models (8) suggested a positive relation between FSP participation and BMI and obesity in nonelderly adult women, although not in men. However, the magnitude of the associations among women varied widely depending on the data set and the analytic methods used, and many studies used data collected before 2000. Longitudinal studies using dynamic models found no association between FSP participation and BMI (9, 10). In one of these studies, food insecurity was positively related to BMI among elderly nonparticipants, but was not related to BMI among elderly FSP participants (10). Studies using instrumental variable models have also yielded inconsistent results (5, 11), likely owing to the difficulty in selecting a valid instrument (12).
If a causal relation exists between SNAP participation and BMI, the mechanisms are potentially complicated. The first hypothesized mechanism is that the association may be mediated by dietary behaviors, particularly consumption of inexpensive, energy-dense foods of minimal nutritional value (13–15). Wilde et al (16) observed that FSP participants consumed more meats, added sugars, and total fats as a result of program participation. Over time, these foods may contribute to higher total energy intake among program participants (4). The results have been inconclusive among other studies examining SNAP participation and dietary intake. The second hypothesized mechanism relates to the monthly distribution of SNAP benefits, and the availability of benefits affecting food purchasing and dietary behaviors throughout the month (17). Participants may overcompensate for energy intake during the weeks when SNAP benefits are available and restrict food intake when SNAP benefits are depleted (17). This cyclic behavior may lead to weight gain over time (18). However, more research is needed to determine how the frequency of benefits distribution affects dietary intake and weight management among low-income SNAP participants.
Obesity is a strong risk factor for several health conditions, including hypertension, hypercholesterolemia, and hyperglycemia. Collectively, these conditions, known as the metabolic syndrome, increase the risk of coronary heart disease, stroke, and diabetes mellitus (19). Whereas the associations between SNAP participation and BMI have been examined in previous studies, little is known about the relations between SNAP participation and other adiposity measures or metabolic risk factors.
Using a nationally representative sample of nonelderly low-income adults, we examined the associations of SNAP participation with BMI and waist circumference, independent of sociodemographic characteristics and household food insecurity. We also examined whether SNAP participants and nonparticipants had different prevalences of metabolic risk factors after sociodemographic characteristics, household food insecurity, and BMI were accounted for and whether the associations differed between men and women.
SUBJECTS AND METHODS
Study population
NHANES is an ongoing, multistage cross-sectional survey administered by the National Center for Health Statistics. NHANES is designed to be representative of the civilian, noninstitutionalized US population and collects information on general health status, nutritional intake, health-related behaviors, and physiologic measurements from an in-home questionnaire and a physical examination in the Mobile Examination Centers. Laboratory analysis of blood and urine samples are also conducted on a subset of NHANES participants.
This analysis combined data from the 2003–2004 and 2005–2006 surveys to ensure a sufficient representation of SNAP participants. The analytic sample was restricted to households whose incomes fell at or below 130% of the federal poverty level to include only individuals who may be eligible to receive SNAP benefits. We further restricted our analysis to adults aged 18–65 y and excluded pregnant women to avoid misclassification of our outcome measures. The sample consisted of 2250 adults; however, the sample size varied across analytic models as certain outcomes were only collected among a subset of study participants.
Outcomes
Outcome variables for analyses included BMI, waist circumference, plasma fasting triglycerides, HDL cholesterol, blood pressure, fasting glucose, and the metabolic syndrome.
Adiposity measurements
Height (in m) and weight (in kg) were measured by trained personnel with a Toledo weight scale and stadiometer (20). BMI was calculated as weight (kg) divided by the square of height (m) and further collapsed into standard weight categories: underweight (BMI <18.5), normal weight (BMI 18.5–24.9), overweight (BMI 25.0–29.9), and obese (BMI ≥30.0) (21). Waist circumference (cm) was measured by trained personnel at the upper lateral border of the right ilium (20). For analysis as an adiposity outcome, waist circumference was categorized by sex-specific quartiles. For men, quartiles were defined at 82.5, 93.25, and 103.8 cm. For women, quartiles were defined at 81.8, 93.0, and 105.75 cm.
Metabolic risk factors
The National Cholesterol Education Program ATP III guidelines were used to define the cutoffs for the metabolic risk factors (elevated waist circumference, elevated triglycerides, lower HDL cholesterol, and elevated blood pressure) (22). Elevated waist circumference was defined as ≥102 cm for men and ≥88 cm for women. Plasma triglycerides (mg/dL) were measured in the morning from persons who had fasted for ≥8.5 h Details of the laboratory procedures are described elsewhere (23). Elevated triglycerides were defined as ≥150 mg/dL. Serum HDL cholesterol (mg/dL) was measured from blood specimens obtained from the mobile examination center laboratory. Details of specimen collection and processing are described elsewhere (24). Lower HDL cholesterol was defined as <40 mg/dL for men and <50 mg/dL for women. For blood pressure, 4 consecutive systolic and diastolic blood pressure readings (mm Hg) were taken in the mobile examination center by certified physical examiners. The average of the first 3 readings was obtained and used for analysis. For individuals with <3 readings, the average of the first and/or second readings was used for analysis. Individuals were excluded if they had partial or missing blood pressure status, or reported consuming food, alcohol, cigarettes, or coffee within the past 30 min. Elevated blood pressure was defined as ≥130/85 mm Hg. Fasting plasma glucose (mg/dL) was measured in the morning from persons who had fasted for ≥9 h Details of the laboratory procedures are described elsewhere (25). Because there has been disagreement about the optimal threshold for fasting glucose (26), elevated glucose was first defined as ≥100 mg/dL on the basis of the current ADA cutoff for impaired fasting glucose (27) and second as ≥110 mg/dL on the basis of the WHO cutoff for impaired fasting glucose (28). The ADA cutoff was used as a lower boundary for identifying individuals with impaired fasting glucose; the WHO cutoff was used as a more “sensitive” indicator of impaired fasting glucose and to maintain consistency with the classification of metabolic syndrome.
Metabolic syndrome
Per the ATP III guidelines, the metabolic syndrome was defined as the presence of ≥3 of the following conditions: elevated waist circumference, triglycerides, blood pressure, and fasting glucose (≥110 mg/dL) and lower HDL cholesterol (22).
SNAP participation
SNAP participation was defined as answering “yes” to the survey question, “In the last 12 months, were you or any members of your household authorized to receive Food Stamps?” SNAP participation was measured at the household level.
Sociodemographic characteristics
Covariates for multivariate regression analyses included age, sex, race (non-Hispanic white, Hispanic, African American, or other/multiple races), place of birth (United States–born or foreign-born), education level (<12 y, high school graduate or equivalent, some college or Associate's degree, or Bachelor's degree or higher), marital status (single; married or living with partner; widowed or separated or divorced), household size, smoking status (never smoker, former smoker, or current smoker), health insurance status (private insurance, public insurance, or no insurance), poverty income ratio (0–25% FPL, 25.1–50% FPL, 50.1–75% FPL, 75.1–100% FPL, or 100.1–130% FPL), household participation in WIC in the past 12 mo, and household food security (full food security, marginal food security, low food security, or very low food security). Household food security was assessed by using the 18-item US Food Security Survey Module (29).
Statistical analysis
Complex survey weights were used to account for different sampling probabilities and participation rates for the various components of NHANES and to obtain effect estimates and SEs representative of the US population. Weights from the MEC and the FAS were recalculated to reflect the probability of being sampled in the 4-y period (30). MEC weights were used for all models fit for measures of adiposity, HDL cholesterol, and blood pressure. Models for triglycerides, glucose, and the metabolic syndrome used FAS weights. All models were restricted to our study population of interest. SEs were estimated by using robust sandwich variances.
We first compared sociodemographic characteristics and food insecurity between SNAP participants and nonparticipants by using chi-square tests. Next, we evaluated associations between SNAP participation and measures of adiposity by fitting multinomial logistic regression models for the outcomes of BMI categories and waist circumference quartiles. Individuals with a normal BMI (18.5–24.9) and in the first quartile of waist circumference were used as the reference groups. The first model adjusted for age. The multivariate model adjusted for age, sex, race, place of birth, education level, marital status, household size, smoking status, health insurance status, poverty income ratio, WIC participation, and household food security. We also conducted trend tests by running multivariate logistic regression models for the outcome of SNAP participation by using BMI categories and waist circumference quartiles as ordinal variables.
To examine the cross-sectional associations between SNAP participation and metabolic risk factors, Poisson regression models were used to estimate PRs for SNAP participation and the different outcomes. Previous studies have suggested that Poisson regression models with robust variance estimation provide PRs that are consistent, more interpretable, and more conservative than the prevalence ORs estimated from binomial regression in cross-sectional studies (31–33). The first model adjusted for age. The multivariate model included all sociodemographic characteristics and household food security as covariates. The third model added BMI categories to assess the relations between SNAP participation and metabolic risk factors independent of the effects from BMI; a missing indicator was used to account for adults with missing BMI data (n = 97).
We also examined whether the associations between SNAP and measures of adiposity or metabolic risk factors were modified by sex. Regression models were fit for each outcome separately for men and women. All statistical tests were 2-sided, and significance was considered at P < 0.05. Statistical analyses were performed by using Stata/IC 11.1 (StataCorp LP).
RESULTS
Of the 2250 adults in the study population, 32.8% reported participation in SNAP within the past 12 mo. Compared with nonparticipants, SNAP participants were more likely to be female, to have been born in the United States, to be African American, to have <12 y of formal education, to be living in a larger household, and to be insured with public types of health insurance (eg, Medicaid, state-sponsored health plan, or other government insurance) (Table 1). Household food insecurity was prevalent among SNAP participants; 24.6% experienced low food security and 21.3% experienced very low food security in the past year.
TABLE 1.
Sociodemographic characteristics of adults at or below 130% of the federal poverty level1
| SNAP participants (n = 723) | SNAP nonparticipants (n = 1527) | |
| Age (y) | 36.3 ± 0.72 | 36.9 ± 0.8 |
| Female [n (%)] | 402 (59.1) | 776 (51.0) |
| Born in the United States [n (%)] | 598 (83.7) | 957 (70.8) |
| Race [n (%)] | ||
| Non-Hispanic white | 222 (47.6) | 467 (52.4) |
| African American | 297 (30.0) | 350 (13.7) |
| Hispanic/Latino | 183 (19.0) | 626 (26.4) |
| Other or multiple ethnicities | 21 (3.5) | 84 (7.4) |
| Education level [n (%)] | ||
| <12 y | 351 (43.3) | 603 (31.8) |
| High school graduate or equivalent | 191 (29.0) | 420 (29.0) |
| Some college or AA degree | 154 (24.8) | 406 (31.0) |
| College graduate or above | 17 (2.9) | 78 (8.2) |
| Marital status [n (%)] | ||
| Single | 268 (29.7) | 625 (32.6) |
| Married or living with partner | 290 (43.9) | 656 (49.2) |
| Widowed, separated, or divorced | 165 (26.4) | 245 (18.2) |
| Household size (n) | 3.8 ± 0.1 | 3.5 ± 0.1 |
| Smoking status [n (%)] | ||
| Never smoker | 259 (34.7) | 743 (49.4) |
| Former smoker | 94 (15.7) | 182 (14.1) |
| Current smoker | 355 (49.6) | 562 (36.5) |
| Health insurance status [n (%)] | ||
| Not insured | 240 (31.7) | 722 (47.1) |
| Insured with public insurance | 404 (57.1) | 343 (22.0) |
| Insured with private insurance | 75 (11.2) | 441 (31.9) |
| Poverty income ratio [n (%)] | ||
| 0–25% of FPL | 93 (12.1) | 215 (12.6) |
| 25.1–50% of FPL | 128 (15.6) | 160 (9.7) |
| 50.1–75% of FPL | 177 (24.6) | 270 (16.0) |
| 75.1–100% of FPL | 185 (27.4) | 397 (24.5) |
| 100.1–130% of FPL | 140 (20.2) | 485 (37.1) |
| Household WIC participation [n (%)] | 244 (32.2) | 263 (15.2) |
| Household food security [n (%)] | ||
| Full food security | 272 (37.4) | 848 (58.4) |
| Marginal food security | 117 (16.6) | 223 (12.9) |
| Low food security | 188 (24.6) | 319 (19.1) |
| Very low food security | 145 (21.3) | 136 (9.6) |
All differences in sociodemographic characteristics between SNAP participants and SNAP nonparticipants, except for age, were statistically significant, P < 0.05 (chi-square tests). AA, Associates of Arts; FPL, federal poverty level; SNAP, Supplemental Nutrition Assistance Program; WIC, Special Supplemental Nutrition Program for Women, Infants and Children.
Mean ± SE (all such values).
Among SNAP participants, 23.7% were overweight and 44.0% were obese (Table 2). After adjustment for sociodemographic characteristics, the odds of obesity was 58% higher among SNAP participants than among nonparticipants (OR: 1.58; 95% CI: 1.08, 2.31; P-trend = 0.006). There was no significant effect modification by gender of the association between SNAP participation and BMI/obesity. SNAP participation was also positively associated with waist circumference in men and women. For men, SNAP participants had twice the odds of being in the top quartile of waist circumference compared with male nonparticipants (OR = 2.04, 95% CI: 1.15, 3.62; P-trend = 0.06). For women, SNAP participants had nearly 3 times the odds of being in the top quartile of waist circumference compared with female nonparticipants (OR = 2.95, 95% CI: 1.51, 5.77; P-trend = 0.001). Despite the difference in the associational measures, the interaction was not statistically significant (P-interaction with sex = 0.11).
TABLE 2.
Participation in SNAP and associations with adiposity among adults at or below 130% of the federal poverty level1
| Age-adjusted2 |
Multivariate-adjusted |
|||||
| SNAP participants | SNAP nonparticipants | OR | 95% CI | OR | 95% CI | |
| n (%) | n (%) | |||||
| BMI | ||||||
| Normal, 18.5–24.9 kg/m2 | 221 (30.1) | 539 (36.9) | Ref | Ref | ||
| Underweight, <18.5 kg/m2 | 15 (2.3) | 43 (3.5) | 0.93 | 0.48, 1.79 | 0.58 | 0.26, 1.28 |
| Overweight, 25.0–29.9 kg/m2 | 169 (23.7) | 457 (33.1) | 0.90 | 0.67, 1.19 | 0.81 | 0.59, 1.10 |
| Obese, ≥30 kg/m2 | 295 (44.0) | 414 (26.6) | 2.15 | 1.63, 2.85 | 1.58 | 1.08, 2.31 |
| P-trend | <0.001 | 0.006 | ||||
| Waist circumference, men only | ||||||
| Quartile 1, <82.5 cm | 71 (15.1) | 178 (21.8) | Ref | Ref | ||
| Quartile 2, 82.5 to <93.25 cm | 76 (27.1) | 178 (24.9) | 1.69 | 0.97, 2.92 | 1.87 | 0.93, 3.74 |
| Quartile 3, 93.25 to <103.8 cm | 73 (26.0) | 177 (28.5) | 1.49 | 0.81, 2.74 | 1.79 | 0.87, 3.68 |
| Quartile 4, ≥103.8 cm | 85 (31.9) | 168 (24.8) | 2.17 | 1.29, 3.64 | 2.04 | 1.15, 3.62 |
| P-trend | 0.02 | 0.06 | ||||
| Waist circumference, women only | ||||||
| Quartile 1, <81.8 cm | 66 (16.8) | 203 (26.4) | Ref | Ref | ||
| Quartile 2, 81.8 to <93.0 cm | 83 (23.4) | 192 (30.9) | 1.16 | 0.64, 2.08 | 1.42 | 0.70, 2.89 |
| Quartile 3, 93.0 to <105.75 cm | 95 (22.5) | 184 (24.6) | 1.44 | 0.87, 2.38 | 1.33 | 0.66, 2.68 |
| Quartile 4, ≥105.75 cm | 131 (37.3) | 143 (18.1) | 3.68 | 2.13, 6.37 | 2.95 | 1.51, 5.77 |
| P-trend | <0.001 | 0.001 | ||||
Ref, reference; SNAP, Supplemental Nutrition Assistance Program.
ORs and 95% CIs were estimated from multinomial logistic regression models with normal BMI and quartile 1 of waist circumference as reference groups.
Adjusted for age; sex; race; place of birth; education level; marital status; household size; health insurance status; poverty income ratio; smoking status; Special Supplemental Nutrition Program for Women, Infants and Children participation; and household food security. Sex was excluded as a covariate in waist circumference models specific to men and women.
We next examined whether SNAP participation was associated with metabolic risk factors (Table 3). Among SNAP participants, the most prevalent risk factors were elevated waist circumference, elevated fasting triglycerides, and lower HDL cholesterol. After adjustment for sociodemographic characteristics and household food security, SNAP participation was associated with elevated triglycerides (PR: 1.71; 95% CI: 1.33, 2.20) and lower HDL cholesterol (PR: 1.23; 95% CI: 1.08, 1.41). After the effect of BMI was accounted for, the associations between SNAP participation and elevated triglycerides and lower HDL cholesterol remained significant. Although the association with elevated triglycerides appeared stronger for women (PR: 1.97; 95% CI: 1.29, 3.02) than for men (PR: 1.45; 95% CI: 0.99, 2.11; P = 0.06) (see Tables S1 and S2 under “Supplemental data” in the online issue), the interaction was not statistically significant (P = 0.26).
TABLE 3.
Participation in the SNAP and associations with metabolic risk factors among adults at or below 130% of the federal poverty level1
| Age-adjusted2 |
Multivariate-adjusted23 |
Multivariate- + BMI-adjusted |
||||||
| SNAP participants | SNAP nonparticipants | PR | 95% CI | PR | 95% CI | PR | 95% CI | |
| n (%) | n (%) | |||||||
| Waist circumference | ||||||||
| Normal, <102 cm for M; <88 cm for F | 323 (44.8) | 814 (56.6) | Ref | Ref | Ref | |||
| Elevated, ≥102 cm for M; ≥88 cm for F | 357 (55.2) | 609 (43.4) | 1.29 | 1.15, 1.45 | 1.14 | 0.99, 1.31 | 0.99 | 0.88, 1.10 |
| Triglycerides | ||||||||
| Normal, <150 mg/dL | 200 (62.8) | 480 (72.4) | Ref | Ref | Ref | |||
| Elevated, ≥150 mg/dL | 111 (37.3) | 171 (27.6) | 1.45 | 1.13, 1.84 | 1.71 | 1.33, 2.20 | 1.67 | 1.28, 2.18 |
| HDL | ||||||||
| Normal, ≥40 mg/dL for M; ≥50 mg/dL for F | 426 (59.4) | 962 (67.7) | Ref | Ref | Ref | |||
| Lower, <40 mg/dL for M; <50 mg/dL for F | 256 (40.6) | 431 (32.3) | 1.26 | 1.13, 1.40 | 1.23 | 1.08, 1.41 | 1.16 | 1.00, 1.36 |
| Blood pressure4 | ||||||||
| Normal, <130/85 mm Hg | 298 (77.2) | 672 (75.3) | Ref | Ref | Ref | |||
| Elevated, ≥130/85 mm Hg | 89 (22.8) | 195 (24.7) | 0.90 | 0.73, 1.12 | 0.93 | 0.73, 1.19 | 0.90 | 0.68, 1.18 |
| Fasting glucose, ADA cutoff | ||||||||
| Normal, <100 mg/dL | 204 (67.1) | 433 (65.8) | Ref | Ref | Ref | |||
| Elevated, ≥100 mg/dL | 112 (32.9) | 225 (34.2) | 1.00 | 0.80, 1.24 | 1.18 | 0.90, 1.55 | 1.08 | 0.81, 1.45 |
| Fasting glucose, WHO cutoff | ||||||||
| Normal, <110 mg/dL | 262 (88.4) | 568 (84.5) | Ref | Ref | Ref | |||
| Elevated, ≥110 mg/dL | 54 (11.6) | 90 (15.5) | 1.39 | 0.96, 2.00 | 1.63 | 1.05, 2.52 | 1.49 | 0.90, 2.47 |
| Metabolic syndrome | ||||||||
| <3 of above conditions present | 265 (73.8) | 659 (81.3) | Ref | Ref | Ref | |||
| ≥3 of above conditions present | 77 (26.2) | 123 (18.7) | 1.51 | 1.09, 2.10 | 1.49 | 1.13, 1.95 | 1.32 | 0.95, 1.81 |
ADA, American Diabetes Association; F, females; M, males; PR, prevalence ratio; Ref, reference; SNAP, Supplemental Nutrition Assistance Program.
PRs and 95% CIs were estimated from Poisson regression models,
Adjusted for age (in 5-y categories); sex; race; place of birth; education level; marital status; household size; health insurance status; poverty income ratio; smoking status; Special Supplemental Nutrition Program for Women, Infants and Children participation; and household food security.
Systolic and diastolic blood pressure calculated as the average of the first 3 readings.
SNAP participation was not significantly associated with elevated fasting glucose on the basis of the ADA cutoff of 100 mg/dL; however, it was associated with the WHO cutoff of 110 mg/dL (PR: 1.63; 95% CI: 1.05, 2.52). This association was attenuated after adjustment for BMI (PR: 1.49; 95% CI: 0.90, 2.47; P = 0.12). The association with elevated fasting glucose (≥110 mg/dL) was stronger for women (PR: 2.03; 95% CI: 1.07, 3.88) than for men (PR: 1.30; 95% CI: 0.66, 2.56); however, the interaction was not significant (P = 0.33).
Approximately 33.6% of SNAP participants met the diagnostic criteria for the metabolic syndrome. Consistent with the WHO and ATP III guidelines, the fasting glucose cutoff of 110 mg/dL was used to define elevated fasting glucose as one criterion for the metabolic syndrome. SNAP participation was positively associated with the metabolic syndrome (PR: 1.49; 95% CI: 1.13, 1.95), although this association was attenuated after adjustment for BMI (PR: 1.32; 95% C:I 0.95, 1.81; P = 0.09). No significant effect modification by sex was found between SNAP participation, the metabolic syndrome, or any other metabolic risk factors (P > 0.20 for all).
DISCUSSION
In this nationally representative sample of nonelderly, low-income adults, we found that household participation in SNAP was associated with obesity, waist circumference, elevated triglycerides, lower HDL cholesterol, elevated fasting glucose (≥110 mg/dL), and the metabolic syndrome. These associations were independent of sociodemographic factors and household food insecurity.
Our findings corroborate results from previous studies that have observed positive relations between SNAP participation and BMI (5, 6, 14, 34–36) and extend the association to waist circumference. When men and women were analyzed separately, our results suggest that the association with waist circumference quartiles was stronger among low-income women, although a positive association was also found among low-income men. SNAP participation was also associated with the prevalence of elevated triglycerides and lower HDL cholesterol and marginally associated with the metabolic syndrome, independent of the effects of BMI. It was previously suggested that associations from cross-sectional studies may be partly explained by reverse causation, ie, overweight or obese individuals are more likely to seek out SNAP benefits because of a desire to consume more food. Because we observed positive associations between SNAP participation and some metabolic risk factors, even after BMI was controlled for, it seems unlikely that reverse causation completely accounts for these associations. These findings may have important public health implications, because previous studies have found that adults with risk factors of the metabolic syndrome are at an increased risk of cardiovascular disease (37) and all-cause mortality (37, 38).
The associations with adiposity and metabolic risk factors may be mediated by dietary behaviors. Studies that examined whether SNAP participants have higher total energy intakes than nonparticipants had mixed results (4, 39–43). A report by Ver Ploeg and Ralston (4) observed that female SNAP participants consumed more calories than did female nonparticipants, but male SNAP participants consumed fewer calories than did male nonparticipants. Aside from total energy, previous studies have observed that SNAP participants consumed significantly more meat products, added sugars, regular soft drinks, total fat, and total energy than did nonparticipants (16, 34, 44). Several of these dietary components have been associated with obesity (45, 46) and the metabolic syndrome among higher-income populations (47–51).
Differential patterns of eating behaviors corresponding to the monthly distribution of SNAP benefits have led researchers to coin the term “the food stamp cycle.” Participants may overcompensate for energy intake during the weeks when SNAP benefits are available and restrict food intake when SNAP benefits are depleted (17). In a 2000 report, mean food expenditure per person per household peaked and declined after the first 3 d of receiving food stamp benefits (52). Throughout the month, individual energy intakes were consistently lower than the Recommended Dietary Allowance for total calories. These averages may have masked differences by household or food type, because it was also observed that the purchase and consumption of specific food items varied depending on the week of benefit availability. Whereas this report suggests that the food stamp cycle may not be a strong contributor to obesity among SNAP participants, more research is needed to confirm the results of this report and to determine the extent that the food stamp cycle may contribute to obesity among certain subgroups of SNAP participants.
A considerable strength of our study was the use of a representative sample of nonelderly, low-income adults. Because anthropometric measures were collected by trained personnel rather than reported by the participants, we were able to minimize potential misclassification of our adiposity outcomes. Furthermore, the NHANES response rates for the 2003–2004 and 2005–2006 surveys were 76% and 77%, respectively (53, 54). These response rates are considerably higher than other state-wide or national health surveys and may help to ensure a representative sample with minimal selection bias.
The key limitation of the current study was its cross-sectional nature, which makes it difficult to infer the temporality of the variables. For example, experiences of food insecurity and SNAP participation were each measured within the previous 12 mo. Because food insecurity levels can change after acquiring SNAP benefits (55), it is difficult to determine whether food insecurity is reported before or after SNAP participation occurred. For this reason, we ran multivariate models, including and excluding household food insecurity as a covariate. The associations did not change significantly with household food insecurity in the model. Thus, we presented associations that adjusted for household food insecurity, because bivariate analyses showed that SNAP participants were more likely to report low food security or very low food security when compared with nonparticipants. Second, some outcome variables had small sample sizes, which may have limited our statistical power to detect associations. Outcomes such as plasma glucose and serum triglycerides were only collected from a subsample of adults who had completed an overnight fast. However, significant associations were observed with these outcomes despite the smaller sample sizes. Furthermore, the NHANES sampling weights adjust for nonresponse specifically of the MEC or the FAS; however, they do not adjust for missing data for our specific outcomes of interest (eg, BMI and blood pressure). For this missingness to introduce bias, missing data would have to be related to both SNAP participation and with metabolic risk factors. In our study, SNAP participation was not related to missingness for any of our outcome variables. Our results were generally robust in various sensitivity analyses. There was also the potential for unmeasured confounding by other lifestyle factors, such as physical activity—a known risk factor for obesity and lower HDL cholesterol—or factors related to SNAP participation, such as length of participation. Although physical activity questionnaires were administered as part of the NHANES 2003–2006, 41% (933/2250) of our study population were missing these variables.
Misclassification of SNAP participation may also exist in our study. SNAP participation is measured at the household level, and we assumed that individuals whose households participate in SNAP share the benefits. Whereas this would generally hold for most household types (eg, households with children, married head households), single persons living together may not always share food costs or consumption and may be incorrectly classified as receiving SNAP. Recent statistics show that only 2.1% of participating households are multiadult households without children, elderly, or disabled persons (56). The prevalence of SNAP participation as estimated from NHANES is lower than the national estimate from the USDA, which suggests that individuals are underreporting SNAP participation. In a previous study, it was hypothesized that men, unmarried individuals, and individuals with higher income are more likely to incorrectly report SNAP participation status, either because of a stigma or lack of knowledge (57). Whereas we have tried to account for the effects of sex, marital status, and poverty income ratio such that, within levels of these variables, the measurement error could be considered nondifferential, it is possible that other differential misclassification of SNAP could bias our results. Future studies might consider correcting for underreporting of SNAP participation by using validation studies.
Researchers have also questioned whether fundamental differences exist between SNAP participants and low-income, eligible nonparticipants that are not captured in large health studies. This has been termed a “selection bias,” where SNAP participants who self-select into public assistance programs possess characteristics that are often unmeasured by national studies. In one study that attempted to account for selection bias, Baum (5) compared estimates from ordinary-least-squares models to those obtained from instrumental variables, fixed-effects, and dynamic models. Significant, positive effects for FSP participation and obesity were observed from ordinary-least-squares models among women; estimates from fixed-effects models were attenuated but still significant. Other results from instrumental variables and dynamic models also supported a positive effect of FSP participation on obesity for women. Findings for men were inconsistent across the models. Baum concluded that, although there was a significant effect of FSP participation on obesity, the effect was small and unlikely to have a significant effect on the national prevalence of obesity. Other studies have also controlled for selection bias using longitudinal models; most studies support a positive association for women, but not for men (4). Although we could not correct for selection bias given the nature of the NHANES data, our results support conclusions from longitudinal studies correcting for selection bias (5–8) and other cross-sectional studies (34, 35). Furthermore, we extend the associations to other adiposity measures and metabolic risk factors in a nationally representative sample of low-income adults.
Recent statistics indicate that the national poverty rate has increased dramatically in the past decade (58). Because SNAP and other food assistance programs act as safety nets for individuals living close to the poverty line, it is important that these programs promote healthy dietary behaviors of their beneficiaries. Whereas longitudinal studies with strong measurement tools are still needed to determine whether these associations are truly causal, programmatic changes to improve the nutritional aspects of the SNAP program may help to improve the health profiles of participants with respect to dietary behaviors.
Acknowledgments
We thank Lilian Cheung for her constructive review of the manuscript.
The authors’ responsibilities were as follows—CWL and WCW: designed the research; CWL and ELD: performed the statistical analysis; and CWL: wrote the manuscript and had primary responsibility for the final content. All authors read and approved the final manuscript. None of the authors had a conflict of interest to declare.
Footnotes
Abbreviations used: ADA, American Diabetes Association; ATP III, Adult Treatment Panel III; FAS, fasting subsample; FPL, federal poverty level; FSP, Food Stamp Program; MEC, mobile examination center; PR, prevalence ratio; SNAP, Supplemental Nutrition Assistance Program; WIC, the Special Supplemental Nutrition Program for Women, Infants and Children.
REFERENCES
- 1.USDA Eligibility. Supplemental Nutrition Assistance Program: Food and Nutrition Service. Washington, DC: USDA, 2009 [Google Scholar]
- 2.USDA Supplemental Nutrition Assistance Program participation and costs. Washington, DC: Food and Nutrition Service, USDA, 2010 [Google Scholar]
- 3.USDA Current participation SNAP. Washington, DC: Food and Nutrition Service, US Department of Agriculture, 2010 [Google Scholar]
- 4.Ver Ploeg M, Ralston K. Food Stamps and obesity: what do we know?. Washington, DC: US Department of Agriculture, Economic Research Service, 2008. EIB-34 [Google Scholar]
- 5.Baum C. The effects of Food Stamps on obesity. Murfreesboro, TN: Middle Tennessee State University, 2007.. (Contractor and Cooperator Report, no. 34.) [Google Scholar]
- 6.Gibson D. Food Stamp Program participation is positively related to obesity in low income women. J Nutr 2003;133:2225–31 [DOI] [PubMed] [Google Scholar]
- 7.Meyerhoefer CD, Pylypchuk Y. Does participation in the Food Stamp Program increase the prevalence of obesity and health care spending? Am J Agric Econ 2008;90:287–305 [Google Scholar]
- 8.Chen Z, Yen ST, Eastwood DB. Effects of Food Stamp Participation on body weight and obesity. Am J Agric Econ 2005;87:1167–73 [Google Scholar]
- 9.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 [DOI] [PubMed] [Google Scholar]
- 10.Kim K, Frongillo EA. Participation in food assistance programs modifies the relation of food insecurity with weight and depression in elders. J Nutr 2007;137:1005–10 [DOI] [PubMed] [Google Scholar]
- 11.Kaushal N. Do food stamps cause obesity? Evidence from immigrant experience. J Health Econ 2007;26:968–91 [DOI] [PubMed] [Google Scholar]
- 12.Hernán MA, Robins JM. Instruments for causal inference: an epidemiologist's dream? Epidemiology 2006;17:360–72 [DOI] [PubMed] [Google Scholar]
- 13.Drewnowski A, Specter SE. Poverty and obesity: the role of energy density and energy costs. Am J Clin Nutr 2004;79:6–16 [DOI] [PubMed] [Google Scholar]
- 14.Zagorsky JL, Smith PK. Does the U.S. Food Stamp Program contribute to adult weight gain? Econ Hum Biol 2009;7:246–58 [DOI] [PubMed] [Google Scholar]
- 15.Ariely D. Predictably irrational: the hidden forces that shape our decisions. New York, NY: HarperCollins Publishers, 2008 [Google Scholar]
- 16.Wilde PE, McNamara PE, Ranney CK. The Effect on dietary quality of participation in the Food Stamp and WIC Programs. Washington, DC: Economic Research Service, USDA, 2000. (Food Assistance and Nutrition Research Report no. 9.) [Google Scholar]
- 17.Dinour LM, Bergen D, Yeh MC. The food insecurity-obesity paradox: a review of the literature and the role food stamps may play. J Am Diet Assoc 2007;107:1952–61 [DOI] [PubMed] [Google Scholar]
- 18.Kroke A, Liese AD, Schulz M, Bergmann MM, Klipstein-Grobusch K, Hoffman K, Boeing H. Recent weight changes and weight cycling as predictors of subsequent two year weight change in a middle-aged cohort. Int J Obes Relat Metab Disord 2002;26:403–9 [DOI] [PubMed] [Google Scholar]
- 19.Grundy SM. Obesity, metabolic syndrome, and cardiovascular disease. J Clin Endocrinol Metab 2004;89:2595–600 [DOI] [PubMed] [Google Scholar]
- 20.CDC Anthropometry procedures. National Health and Nutrition Examination Survey. Atlanta, GA: CDC, 2004 [Google Scholar]
- 21.CDC About BMI for adults. Healthy weight. Atlanta, GA: CDC, 2009 [Google Scholar]
- 22.Grundy SM, Brewer HB, Jr, Cleeman JI, Smith SC, Jr, Lenfant C. Definition of metabolic syndrome: Report of the National Heart, Lung, and Blood Institute/American Heart Association conference on scientific issues related to definition. Circulation 2004;109:433–8 [DOI] [PubMed] [Google Scholar]
- 23.CDC Triglyceride, LDL-cholesterol and apolipoprotein (ApoB). National Health and Nutrition Examination Survey 2005-2006. Atlanta, GA: CDC, 2008 [Google Scholar]
- 24.CDC HDL cholesterol. National Health and Nutrition Examination Survey 2005-2006. Atlanta, GA: CDC, 2007 [Google Scholar]
- 25.CDC Plasma fasting glucose and insulin. National Health and Nutrition Examination Survey 2005-2006. Atlanta, GA: CDC, 2008 [Google Scholar]
- 26.Forouhi NG, Balkau B, Borch-Johnsen K, Dekker J, Glumer C, Qiao Q, Spijkerman A, Stolk R, Tabac A, Wareham NJ. The threshold for diagnosing impaired fasting glucose: a position statement by the European Diabetes Epidemiology Group. Diabetologia 2006;49:822–7 [DOI] [PubMed] [Google Scholar]
- 27.Genuth S, Alberti KG, Bennett P, Buse J, Defronzo R, Kahnh R, Kitzmiller J, Knowler WC, Lebovitz H, Lernmark A, et al. Follow-up report on the diagnosis of diabetes mellitus. Diabetes Care 2003;26:3160–7 [DOI] [PubMed] [Google Scholar]
- 28.WHO Definition and diagnosis of diabetes mellitus and intermediate hyperglycemia. Report of a WHO/IDF Consultation. Geneva, Switzerland: World Health Organization, 2006 [Google Scholar]
- 29.Bickel G, Nord M, Price C, Hamilton W, Cook J. Guide to measuring household food security. Washington, DC: Food and Nutrition Service,USDA, 2000 [Google Scholar]
- 30.Analytic and Reporting Guidelines, The National Health and Nutrition Examination Surveys (NHANES). Hyattsville, MD: National Center for Health Statistics, CDC, 2005 [Google Scholar]
- 31.Barros AJ, Hirakata VN. Alternatives for logistic regression in cross-sectional studies: an empirical comparison of models that directly estimate the prevalence ratio. BMC Med Res Methodol 2003;3:21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Spiegelman D, Hertzmark E. Easy SAS calculations for risk or prevalence ratios and differences. Am J Epidemiol 2005;162:199–200 [DOI] [PubMed] [Google Scholar]
- 33.Thompson ML, Myers JE, Kriebel D. Prevalence odds ratio or prevalence ratio in the analysis of cross sectional data: what is to be done? Occup Environ Med 1998;55:272–7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Leung CW, Villamor E. Is participation in food and income assistance programmes associated with obesity in California adults? Results from a state-wide survey. Public Health Nutr 2010;14:645–52 [DOI] [PubMed] [Google Scholar]
- 35.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 [DOI] [PubMed] [Google Scholar]
- 36.Webb AL, Schiff A, Currivan D, Villamor E. Food Stamp Program participation but not food insecurity is associated with higher adult BMI in Massachusetts residents living in low-income neighbourhoods. Public Health Nutr 2008;11:1248–55 [DOI] [PubMed] [Google Scholar]
- 37.Ford ES. Risks for all-cause mortality, cardiovascular disease, and diabetes associated with the metabolic syndrome: a summary of the evidence. Diabetes Care 2005;28:1769–78 [DOI] [PubMed] [Google Scholar]
- 38.Malik S, Wong ND, Franklin SS, Kamath TV, L'Italien GJ, Pio JR, Williams GR. Impact of the metabolic syndrome on mortality from coronary heart disease, cardiovascular disease, and all causes in United States adults. Circulation 2004;110:1245–50 [DOI] [PubMed] [Google Scholar]
- 39.Butler JS, Raymond JE. The effect of the Food Stamp Program on nutrient intake. Econ Inq 1996;34:781–98 [Google Scholar]
- 40.Rose D, Habicht JP, Devaney B. Household participation in the Food Stamp and WIC programs increases the nutrient intakes of preschool children. J Nutr 1998;128:548–55 [DOI] [PubMed] [Google Scholar]
- 41.Whitfield RA. A nutritional analysis of the Food Stamp Program. Am J Public Health 1982;72:793–9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Devaney B, Moffitt R. Dietary effects of the Food Stamp Program. Am J Agric Econ 1991;73:202–11 [Google Scholar]
- 43.Parks JC, Smith AD, Alston JM. The effects of the Food Stamp Program on energy balance and obesity. Australian Agricultural & Resource Economics Society 2011 Annual Conference. Melbourne, Australia, 31 January, 2011 [Google Scholar]
- 44.Cason KL, Cox RH, Burney JL, Poole KP, Wenrich TR. Do food stamps without education improve the nutrient intake of recipients? Topics Clin Nutr 2002;17:74–82 [Google Scholar]
- 45.Hu FB, Manson JE, Willett WC. Types of dietary fat and risk of coronary heart disease: a critical review. J Am Coll Nutr 2001;20:5–19 [DOI] [PubMed] [Google Scholar]
- 46.Malik VS, Schulze MB, Hu FB. Intake of sugar-sweetened beverages and weight gain: a systematic review. Am J Clin Nutr 2006;84:274–88 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Dhingra R, Sullivan L, Jacques PF, Wang TJ, Fox CS, Meigs JB, D'Agostino RB, Gaziano JM, Vasan RS. Soft drink consumption and risk of developing cardiometabolic risk factors and the metabolic syndrome in middle-aged adults in the community. Circulation 2007;116:480–8 [DOI] [PubMed] [Google Scholar]
- 48.Fogli-Cawley JJ, Dwyer JT, Saltzman E, McCullough ML, Troy LM, Meigs JB, Jacques PF. The 2005 Dietary Guidelines for Americans and risk of the metabolic syndrome. Am J Clin Nutr 2007;86:1193–201 [DOI] [PubMed] [Google Scholar]
- 49.Lutsey PL, Steffen LM, Stevens J. Dietary intake and the development of the metabolic syndrome: the Atherosclerosis Risk in Communities study. Circulation 2008;117:754–61 [DOI] [PubMed] [Google Scholar]
- 50.Millen BE, Pencina MJ, Kimokoti RW, Zhu L, Meigs JB, Ordovas JM, D'Agostino RB. Nutritional risk and the metabolic syndrome in women: opportunities for preventive intervention from the Framingham Nutrition Study. Am J Clin Nutr 2006;84:434–41 [DOI] [PubMed] [Google Scholar]
- 51.Pan Y, Pratt CA. Metabolic syndrome and its association with diet and physical activity in US adolescents. J Am Diet Assoc 2008;108:276–86, discussion 286 [DOI] [PubMed] [Google Scholar]
- 52.Wilde PE, Ranney CK. The monthly Food Stamp cycle: shopping frequency and food intake decisions in an endogenous switching regression framework. Am J Agric Econ 2000;82:200–13 [Google Scholar]
- 53.CDC Unweighted response rates for NHANES 2005-2006 by age and gender. Atlanta, GA: CDC; Available from: http://www.cdc.gov/nchs/nhanes/response_rates_CPS.htm [Google Scholar]
- 54.CDC Unweighted response rates for NHANES 2003-2004 by age and gender. Atlanta, GA: CDC; Available from: http://www.cdc.gov/nchs/nhanes/response_rates_CPS.htm [Google Scholar]
- 55.Nord M, Golla AM. Does SNAP decrease food insecurity? Untangling the self-selection effect. Washington, DC: Economic Research Service, USDA, 2009 [Google Scholar]
- 56.Leftin J, Gothro A, Eslami E. Characteristics of Supplemental Nutrition Assistance Program households: fiscal year 2009. Alexandria, VA: Nutrition Assistance Program Report Series, Office of Research and Analysis, Food and Nutrition Service, USDA, 2010 [Google Scholar]
- 57.Bollinger CR, David MH. Modeling discrete choice with response error: Food Stamp participation. J Am Stat Assoc 1997;92:827–35 [Google Scholar]
- 58.DeNavas-Walt C, Proctor BD, Smith JC. Income, poverty, and health insurance coverage in the United States: 2009. Current Population Reports. Washington, DC: US Census Bureau, US Government Printing Office, 2010:60–238 [Google Scholar]