Gender differences in the association between food insecurity and insulin resistance among U.S. adults: National Health and Nutrition Examination Survey, 2005–2010

Annals of Epidemiology 25 (2015) 643e648

Contents lists available at ScienceDirect

Annals of Epidemiology journal homepage: www.annalsofepidemiology.org

Original article

Gender differences in the association between food insecurity and insulin resistance among U.S. adults: National Health and Nutrition Examination Survey, 2005e2010 Junxiu Liu MD a, *, Yong-Moon Mark Park MD, PhD, MS a, Seth A. Berkowitz MD, MPH b, c, Qingwei Hu MD, PhD a, Kyungdo Han MBioStat d, Andrew Ortaglia PhD, MSPH a, Robert E. McKeown PhD a, Angela D. Liese PhD, MPH a a

Department of Epidemiology and Biostatistics, Arnold School of Public Health, University of South Carolina, Columbia, SC Division of General Internal Medicine, Massachusetts General Hospital, Boston c Harvard Medical School, Boston, MA d Department of Biostatistics, College of Medicine, The Catholic University of Korea, Seoul, South Korea b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 27 October 2014 Accepted 13 June 2015 Available online 19 June 2015

Purpose: To examine gender-specific associations between food insecurity and insulin resistance in a representative U.S. population. Methods: Data on 5533 adults of 20 years of age or more (2742 men and 2791 women) without diabetes from the 2005e2010 National Health and Nutrition Examination Survey were analyzed. Respondents were categorized as having full, marginal, low, or very low food security using a validated scale. Insulinresistant individuals were defined as those with a homeostasis model assessment of insulin resistance value 2.5 or more. Results: Insulin resistance was higher in both normal-weight (P ¼ .001) and overweight or obese (P < .001) women with lower food security, but no linear trend was found in men. In multiple logistic regression analyses, however, very low food securitydcompared with full food securitydwas associated with insulin resistance in normal-weight men (odds ratio, 3.99; 95% confidence interval, 1.71e9.33), and marginal food insecurity was associated with insulin resistance in overweight or obese men (odds ratio, 2.07; 95% confidence interval, 1.18e3.64) after adjusting for potential confounders. In women, the association between food insecurity and insulin resistance was no longer significant after adjustment. Conclusions: Food insecurity is associated with insulin resistance in adults without diabetes, and this effect varies by gender in normal-weight and overweight or obese populations. Improving food security status may help reduce insulin resistance, an underlying risk factor for diabetes and cardiovascular disease. Ó 2015 Elsevier Inc. All rights reserved.

Keywords: Food insecurity Insulin resistance Gender difference NHANES

Introduction Initially conceived in the 1990s, the term “food insecurity” has been defined as “the limited or uncertain availability of nutritional adequate and safe foods or the limited or uncertain ability to J.L. and Y.-M.M.P. contributed equally to this work. S.A.B. was supported by an Institutional National Research Service Award T32HP10251, the Ryoichi Sasakawa Fellowship Fund, and the Division of General Internal Medicine at Massachusetts General Hospital. Otherwise, the authors declare that there are no conflicts of interest. * Corresponding author. Department of Epidemiology and Biostatistics, Arnold School of Public Health, University of South Carolina, 915 Greene Street, Columbia, SC 29208. Tel.: þ1-803-777-7353; fax: þ1-803-777-2524. E-mail address: [email protected] (J. Liu). http://dx.doi.org/10.1016/j.annepidem.2015.06.003 1047-2797/Ó 2015 Elsevier Inc. All rights reserved.

acquire acceptable foods in socially acceptable ways.” [1] In 2011, approximately 15% of all U.S. households (w49 million people) were food insecure [2]. Food insecurity occurs in all low-, middleand high-income countries [1], making it a global public-health burden. Numerous studies have demonstrated the association between food insecurity and poorer health, including increased obesity [3], metabolic syndrome [4], diabetes mellitus [5,6], and several chronic diseases [7]. Food insecure individuals may preferentially purchase inexpensive, energy-dense foods because of their lower cost [8], which can result in increased consumption of saturated fat and refined carbohydrates and decreased consumption of fruits, vegetables, and whole grains. This dietary pattern is associated with insulin resistance [9], type 2 diabetes [10], and cardiovascular disease [11]. In addition, studies have documented

644

J. Liu et al. / Annals of Epidemiology 25 (2015) 643e648

that certain dietary behaviors, such as episodes of underconsumption and overconsumption [12] and substitution of energy-dense foods [13]dcommon coping strategies in food insecuritydmay be associated with insulin resistance [14]. Insulin resistance, a reduction to below-normal levels of insulin’s ability to mediate physiological function, is presumed to be the common underlying pathogenic link among the various components of the metabolic syndrome [15], as well as a risk factor for both diabetes mellitus [16] and coronary heart disease [17]. Even in normal-weight individuals, insulin resistance may explain a portion of cardiovascular disease risk [18]. It has been suggested that not all individuals with normal weight have a healthy metabolic profile [19] and not all obese individuals present with metabolic and cardiovascular risk factors [20]. Thus, it may be hypothesized that the relation between food insecurity and insulin resistance differs by obesity status. In addition, the effects of food insecurity may differ by gender, especially given differently emphasized social roles among men and women, such as childbearing and caregiving [21]. Food insecurity is associated with a range of adverse health outcomes through its influence on both physical and psychosocial factors. As a measure of socially constructed differences on the basis of biological sex, gender might reflect differences in the association between food insecurity and health outcomes [22]. Gender has been strongly suggested by previous studies as a moderator variable between food insecurity and certain health outcomes, such as obesity, depression, and other mental-health issues [23]. In addition, previous research has proposed potential mechanisms that involve biology, behavior, and sociocultural factors as the bases for gender differences in health-related outcomes [24]. Hence, the health consequences of food insecurity might be more pronounced in one gender than the other. Thus, this study was conducted to fill the literature gap in the link between food insecurity and insulin resistance by testing the hypotheses that food insecurity is related to insulin resistance and that this association varies by gender. The study used data from a representative sample of normal-weight and overweight or obese U.S. adults aged 20 years or more in the National Health and Nutrition Examination Survey (NHANES), 2005e2010. Material and methods Study population NHANES is a cross-sectional, nationally representative survey of the noninstitutionalized population of the United States administered by the National Center for Health Statistics [17]. NHANES uses multistage probability cluster sampling to obtain a nationally representative sample of the U.S. population in 2-year cycles. The survey consists of three components: a household interviewer-administered survey, a physical examination, and laboratory tests. Details of recruitment and data collection procedures have been published previously [25]. Participants provided written informed consent, and the National Center for Health Statistics Research Ethics Review Board reviewed and approved this study protocol [26]. Three waves of NHANES (2005e2006, 2007e2008, and 2009e2010) were combined to create an analytic data set using appropriate sample weights to approximate the U.S. noninstitutionalized adult population over the period of the surveys [27]. During these three NHANES waves, 7392 individuals aged 20 years or more completed the food security survey module and the fasting laboratory examinations and were thus eligible for this study. Participants with missing data on body mass index (BMI; n ¼ 109) and those with BMI less than 18.5 kg/m2 (n ¼ 119), pregnancy (n ¼ 217), or diabetes mellitus (n ¼ 1421) were excluded from analysis.

Pregnancy status was ascertained through a urine pregnancy test or self-report and was determined as part of the NHANES protocols in all cycles [28]. Subjects with diabetes mellitus were excluded from analysis because the homeostatic model assessment of insulin resistance (HOMA-IR) correlates well with insulin resistance in a population without diabetes [29] but is not valid for participants with diabetes. Diabetes mellitus was defined as having at least one of the following: self-reported diabetes diagnosis; currently taking insulin or oral hypoglycemic agents; fasting plasma glucose 7.0 mmol/L or more; 2-hour glucose tolerance 11.1 mmol/L or more; or glycated hemoglobin levels 6.5% or more. After applying the exclusion criteria, 5533 participants were included in analysis. Data were not reweighted after exclusions, and following SAS guidelines for the analysis of complex samples, individuals were not deleted from data sets even when they were not included in the analysis. Measurement of food insecurity Food insecurity during the past 12 months is measured at the household level in the NHANES Food Security Survey Module, which is a well-validated instrument developed by the U.S. Department of Agriculture (USDA) [30]. Instead of using the full 18 items from the NHANES module, only the 10 adult-specific household items were used in this analysis. The household food insecurity scale was summarized as a single overall measure based on the set of food security questions in the core survey module. In this study, food security status was categorized according to NHANES responses using standard cutoff points: 0 affirmative responses (full food security), 1e2 affirmative responses (marginal food security), 3e5 affirmative responses (low food security), or 6e10 affirmative responses (very low food security) [30]. Definitions of insulin resistance and body weight status Serum insulin was initially measured using the Mercodia sandwich ELISA assay (Mercodia, Uppsala, Sweden), and the measurement technique was switched in late 2009 to a Roche chemiluminescent immunoassay performed on the Elecsys 2010 analyzer (Roche Diagnostics, Indianapolis, IN). Plasma glucose measurements were performed using the hexokinase assay on the Roche/ Hitachi 911 for NHANES 2005e2006 samples and the Roche Modular P chemistry analyzer for NHANES 2007e2010 samples (Roche Diagnostics). Height and weight of the participants were measured using a fixed stadiometer and a Toledo digital scale (METTLER-TOLEDO Inc., Columbus, OH), respectively [28]. Insulin resistance was estimated using the following formula: HOMA-IR ¼ fasting insulin (l U/mL)  fasting plasma glucose (mM)/ 22.5, where scores greater than 2.5 indicate insulin resistance [31]; this method provides a reliable estimate of insulin resistance that has been validated with the hyperinsulinemic-euglycemic clamp technique [32]. Weight and height were used to calculate BMI as follows: weight in kilograms divided by the square of height in meters. Subjects were classified as normal weight (BMI ¼ 18.5e24.9 kg/m2) or overweight or obese (BMI 25.0 kg/m2). Measurement of covariates Age, sex, race and/or ethnicity (non-Hispanic white, nonHispanic black, Hispanic, or other), educational level (at least a high school degree [>12 years] or not), and poverty-income ratio (PIR: the ratio of a household’s income to the appropriate federal poverty line for the respondent’s household size) were assessed by questionnaire during the household interview. PIR less than 1 indicated low income. Smoking status was defined as ever smoker

J. Liu et al. / Annals of Epidemiology 25 (2015) 643e648

or not based on cigarette use questions, if the participant smoked at least 100 cigarettes in his or her life regardless of current smoking status. Alcohol consumption was dichotomized as consuming at least 12 alcohol drinks in the previous year or not, and physical activity was defined as any consistent moderate or vigorous activity for at least 10 min/d or no such activity over the past 30 days. The 24-hour dietary recall data in day one were used to assess diet and food consumption. These data were collected through in-person interviews during the Mobile Examination Center examination using the Automated Multiple-Pass Method, which was developed by the USDA. Total energy intake (kilocalories) and total fat (gram) in the previous 24 hours were determined from this interview, as estimated for each participant using the USDA database nutrient composition of the foods the participants reported they consumed [33]. Statistical analysis All data were analyzed using SAS, version 9.3 (SAS Institute, Cary, NC), using procedures for complex samples. Continuous variables were described by the mean and standard error (SE) and were compared using linear regression analyses (PROC SURVEYREG [SAS Institute]). Categorical variables were reported by percentage with the SE and were compared using Rao-Scott c2 tests (PROC SURVEYFREQ [SAS Institute]). P values less than .05 were considered statistically significant. All statistical analyses were performed using the appropriate survey procedures to account for complex sampling design and weights. For the subgroup analyses, domain analysis was applied, so that the entire sample was used for estimating the variance of the subpopulations. Multiple logistic regression analyses were conducted to quantify the association between insulin resistance and food insecurity within normal-weight and overweight or obese individuals. The covariates included in the models were age, race and/or ethnicity, smoking status, alcohol consumption, physical activity, education level, PIR, BMI, total energy intake, and total fat intake. Total energy intake and total fat intake as potential confounders were adjusted for because it is usually associated with metabolic disease risk and insulin resistance, respectively. In these multivariable models, BMI was further adjusted as a residual confounder. In addition, effect

645

modification by gender, BMI, total fat intake, total energy intake, and physical activity on the association between food insecurity and insulin resistance was assessed. The specific interaction terms of food insecurity with gender, BMI, total energy intake, total fat intake, and physical activity were evaluated separately to explore potential effect modifications. Results Demographic and clinical characteristics are presented according to food security status in men and women separately (Tables 1 and 2). Overall, participants from households with full food security were more likely to be older, non-Hispanic white, never smokers, and physically active, in addition to having higher income and education, compared with those from households with lower food security. With lower food security, participants’ HOMA-IR and fasting insulin levels were higher, and HDL cholesterol levels were lower in women only (all P < .0001). BMI and waist circumference were also higher with lower food security in women (all P < .0001), but these measures tended to be lower in men. The interaction terms for food insecurity with gender and BMI were statistically significant (P < .05), and the interaction terms for food insecurity with nutrient intake and physical activity were not statistically significant at a ¼ 0.05. Figure 1 shows the prevalence of insulin resistance according to food security status for normal-weight and overweight or obese participants. Overall, the prevalence of insulin resistance in participants with normal weight was highest in those reporting very low food security for both men (33.6%) and women (22.5%). In women, there was an age-adjusted linear trend (P ¼ .001) of higher insulin resistance with greater food insecurity, but this trend was not seen in men. In comparison, in participants who were overweight or obese, the highest prevalence of insulin resistance was observed in men reporting marginal food security status (74.1%) and in women reporting low food security (66.1%). In women, there was again an age-adjusted linear trend (P < .0001) of greater insulin resistance with greater food insecurity, but this trend was not present in men. In multivariable logistic regression models adjusted for the covariates listed previously, very low food security was significantly associated with increased odds of insulin resistance in normal-weight men (odds ratio, 3.99; 95%

Table 1 General characteristics of representative U.S. adult men by household food security status, NHANES, 2005e2010 Characteristic Age (y) Race/ethnicity, % Mexican American Other Hispanic Non-Hispanic white Non-Hispanic black Other BMI (kg/m2) Waist circumference (cm) Fasting plasma glucose (mmol/L) Fasting serum insulin (pmol/L) HOMA-IR Total fat intake (g) Total energy intake (kcal) Education level (>high school degree), % Low income, % Ever smoker, % Regular exercise, % Alcohol consumption, %

Overall (n ¼ 2742) 43.9 (0.5) 9.1 4.7 70.5 9.9 5.8 28.3 99.9 5.58 72.7 3.05 99.5 2639.1 58.0 11.6 57.8 63.7 56.2

(1.0) (0.7) (1.8) (0.9) (0.6) (0.1) (0.4) (0.02) (1.7) (0.07) (1.4) (30.9) (1.8) (0.7) (1.4) (1.6) (2.3)

Full food security (n ¼ 2006) 45.5 (0.6) 6.3 3.9 75.8 7.9 6.2 28.4 100.6 5.58 72.6 3.13 100.6 2629.2 63.3 6.8 50.5 66.6 56.9

(0.8) (0.5) (1.6) (0.8) (0.7) (0.2) (0.5) (0.02) (1.7) (0.08) (1.5) (31.7) (2.1) (0.6) (1.6) (2.0) (2.5)

Marginal food security (n ¼ 263) 36.9 (1.0) 21.3 8.8 49.6 17.3 3.1 28.4 99.4 5.55 78.5 3.59 97.0 2678.5 39.3 26.0 57.1 53.5 55.4

DBP ¼ diastolic blood pressure; HDL ¼ high-density lipoprotein; SBP ¼ systolic blood pressure. Data are presented as the mean (SE) or percentage (SE). 2 * P for race and/or ethnicity was based on RaoeScott c , not for trend.

(3.0) (2.7) (5.0) (3.0) (1.7) (0.5) (1.4) (0.05) (5.4) (0.33) (4.9) (107.7) (4.4) (3.3) (4.0) (4.6) (6.5)

Low food security (n ¼ 319) 37.0 (0.9) 21.9 7.3 48.6 16.8 5.5 27.3 95.7 5.51 68.4 2.92 93.6 2694.8 36.3 32.6 65.1 48.4 51.9

(3.3) (1.9) (5.0) (2.5) (1.6) (0.3) (1.0) (0.03) (3.9) (0.18) (4.7) (126.1) (4.1) (3.7) (3.6) (3.4) (5.3)

Very low food security (n ¼ 154) 37.9 (1.31) 17.1 8.8 48.5 21.5 4.1 27.9 95.9 5.58 74.1 3.12 94.2 2652.8 29.1 39.2 69.2 56.2 54.5

(3.5) (3.4) (6.7) (4.6) (2.6) (1.1) (1.5) (0.04) (5.4) (0.24) (6.3) (149.8) (5.5) (6.9) (5.4) (4.5) (10.6)

P for linear trend <.001 <.001*

.033 <.0001 .217 .392 .389 .319 .937 <.0001 <.0001 <.0001 <.0001 .031

646

J. Liu et al. / Annals of Epidemiology 25 (2015) 643e648

Table 2 General characteristics of representative U.S. adult women by household food security status, NHANES, 2005e2010 Characteristic Age (y) Race/ethnicity, % Mexican American Other Hispanic Non-Hispanic white Non-Hispanic black Other BMI (kg/m2) Waist circumference (cm) Fasting plasma glucose (mmol/L) Fasting serum insulin (pmol/L) HOMA-IR Total fat intake (g) Total energy intake (kcal) Education level (>high school degree), % Low income, % Ever smoking, % Regular exercise, % Alcohol consumption, %

Overall (n ¼ 2791) 46.0 (0.4) 7.0 4.5 71.0 11.3 6.3 28.3 93.8 5.32 64.4 2.49 68.4 1804.1 61.5 12.9 41.0 57.9 69.3

(0.8) (0.7) (1.9) (1.0) (0.8) (0.1) (0.4) (0.02) (1.3) (0.06) (0.8) (17.4) (1.3) (0.9) (1.6) (1.6) (1.6)

Full food security (n ¼ 1996) 47.5 (0.5) 4.8 3.2 76.9 9.0 6.1 27.9 93.0 5.32 60.7 2.47 68.5 1787.8 66.0 8.8 39.3 61.6 70.1

(0.6) (0.5) (1.5) (0.9) (0.9) (0.2) (0.4) (0.02) (1.4) (0.06) (0.1) (21.1) (1.5) (0.8) (1.7) (1.8) (1.7)

Marginal food security (n ¼ 305) 40.4 (1.2) 15.6 10.0 52.2 18.9 3.3 29.5 95.6 5.33 75.5 3.09 70.1 38.8 48.6 22.4 47.2 45.9 65.1

(3.1) (2.5) (5.3) (2.9) (1.5) (0.5) (1.0) (0.05) (4.8) (0.21) (2.2) (0.8) (3.1) (3.2) (3.9) (3.3) (3.8)

Low food security (n ¼ 326) 38.8 (0.8) 17.6 9.5 40.0 21.2 11.7 30.3 98.2 5.31 80.1 3.27 67.5 1853.2 41.1 42.8 45.8 41.0 66.9

(2.7) (2.2) (4.8) (3.1) (2.6) (0.4) (0.8) (0.04) (4.0) (0.18) (2.5) (56.1) (3.2) (4.0) (4.4) (4.0) (3.8)

Very low food security (n ¼ 164) 41.4 (1.5) 12.1 8.7 52.7 21.7 4.8 30.4 97.8 5.35 83.9 3.44 65.2 1864.5 38.5 41.1 51.3 42.2 66.3

(3.2) (2.7) (6.1) (4.3) (2.3) (0.7) (1.6) (0.05) (7.4) (0.32) (2.9) (98.6) (5.9) (5.7) (5.4) (4.9) (6.2)

P for linear trend <.0001 <.0001*

<.0001 <.0001 .910 <.0001 <.0001 .524 .226 <.0001 <.0001 .027 <.0001 .499

DBP ¼ diastolic blood pressure; HDL ¼ high-density lipoprotein; SBP ¼ systolic blood pressure. Data are presented as the mean (SE) or percentage (SE). 2 * P for race and/or ethnicity was based on Rao-Scott c , not for trend.

confidence interval, 1.71e9.33, referent group: full food security); in overweight or obese men, marginal food security was associated with increased odds of insulin resistance (odds ratio, 2.07; 95% confidence interval, 1.18e3.64; Table 3). Low food security was significantly associated with insulin resistance in both normalweight and overweight or obese women after adjusting for age and race and/or ethnicity, but this significant association disappeared after further adjusting for age, race and/or ethnicity, smoking status, alcohol consumption, physical activity, education level, PIR, BMI, total energy intake, and total fat intake. Discussion In a representative U.S. adult population, we observed that higher food insecurity category was associated with increased odds of insulin resistance, but this association varied by gender and BMI. Insulin resistance was associated with very low food security in normal-weight men and with marginal food insecurity in overweight or obese men. In women, insulin resistance was more prevalent with worse food security in both normal-weight and overweight or obese individuals in unadjusted analyses, but this relationship was no longer significant after adjustment. The association between food insecurity and insulin resistance may result from the particular episodic dietary patterns observed in food insecure households, specifically, episodic underconsumption

during times when food is not available and overconsumption during times when food is available [12]. This pattern could be exacerbated by the structure of nutritional assistance programs such as the Supplemental Nutrition Assistance Program, which uses a monthly pay cycle [34]. The stressful state related to food insecurity may also explain the relationship between food insecurity and insulin resistance [35], as it is well known that stress is associated with adiposity, especially visceral obesity, through corticosteroid hormonal reactions, which are closely related to insulin resistance [36]. This study shows that the association between food insecurity and insulin resistance varies by gender and BMI. Findings regarding gender differences in the association between obesity and food insecurity have been inconsistent. A cross-sectional study from the U.S. showed that food insecurity is associated with overweight status in women only [37]. In contrast, Tayie et al. [38] observed an opposite phenomenon: body fat percentage, BMI, and height were lower in men with increasing food insecurity, whereas marginal food security was significantly associated only with height in women. In the present study, with worse food security, obesity parametersdincluding BMI and waist circumferencedwere higher in women, whereas a reverse association was observed in men. These opposite relationships may be explained by different strategies used for coping with food insecurity by gender, such as greater

Fig. 1. Prevalence of insulin resistance according to food security status in normal-weight (A) and overweight or obese (B) populations. The prevalence of insulin resistance increases with worsening food security in normal weight (P for linear trend ¼ .001) and overweight or obese (P for linear trend <.0001) women.

J. Liu et al. / Annals of Epidemiology 25 (2015) 643e648 Table 3 Association between household food insecurity and insulin resistance* according to obese status in U.S. adults, NHANES, 2005e2010 Models

Full food security

Men Overall Model 1 1.00 (ref) Model 2 1.00 (ref) Normal weight Model 1 1.00 (ref) Model 2 1.00 (ref) Overweight/obese Model 1 1.00 (ref) Model 2 1.00 (ref) Women Overall Model 1 1.00 (ref) Model 2 1.00 (ref) Normal weight Model 1 1.00 (ref) Model 2 1.00 (ref) Overweight/obese Model 1 1.00 (ref) Model 2 1.00 (ref)

Marginal food security

Low food security

Very low food security

1.35 (0.97e1.89) 0.86 (0.59e1.24) 1.12 (0.78e1.61) 1.67 (1.08e2.60) 0.91 (0.55e1.52) 1.56 (0.92e2.63) 1.09 (0.41e2.92) 1.17 (0.46e3.00) 3.40 (1.42e8.09) 1.17 (0.38e3.56) 0.65 (0.17e2.46) 3.99 (1.71e9.33) 2.26 (1.40e3.64) 0.87 (0.54e1.39) 0.86 (0.56e1.31) 2.07 (1.18e3.64) 0.95 (0.50e1.79) 0.94 (0.57e1.57)

1.34 (0.96e1.86) 2.14 (1.70e2.70) 1.78 (1.13e2.80) 0.83 (0.51e1.36) 1.35 (0.90e2.01) 1.12 (0.60e2.09) 1.99 (0.76e5.19) 2.28 (1.06e4.89) 2.12 (0.74e6.11) 1.44 (0.55e3.81) 1.99 (0.88e4.54) 1.42 (0.41e4.91) 1.12 (0.75e1.66) 1.72 (1.13e2.62) 1.53 (0.95e2.45) 0.70 (0.43e1.14) 1.19 (0.72e1.97) 1.08 (0.57e2.03)

Data are shown as the odds ratio (95% confidence interval). Adjusted for age and race and/or ethnicity in model 1; additionally adjusted for BMI, smoking status, alcohol consumption, physical activity, education level, poverty-income ratio, total energy intake, and total fat intake in model 2. * Insulin resistance was estimated using the following formula: HOMAIR ¼ fasting insulin (l U/ml)  fasting plasma glucose (mM)/22.5, where scores more than 2.5 indicate insulin resistance.

episodic underconsumption and overconsumption in women than in men [39]. In addition, compared with normal-weight women, overweight or obese women may have different reporting of food insecurity because of different eating habits and greater emphasis on food in their lives [40]. Although the mechanism underlying the observed phenomena remains unclear, sex hormones and social roles may play a part in explaining this association. Cardiometabolic effects of insulin resistance differ by gender [41], which is partially explained by differences in hormonal [42] and lifestyle-related characteristics [43] between men and women. Prior studies have suggested that the effect of estrogen on insulin sensitivity is favorable in women, even though the women in these studies had higher adiposity compared with men [44], and that testosterone mediates the association between unfavorable body fat distribution and metabolic abnormalities in men [45]. In addition, adipokines such as leptin and adiponectin might contribute to gender difference in body composition and insulin resistance [41]. The differences in coping strategies for stress between men and women also may contribute to this association. Stressful states such as food insecurity may be associated with insulin resistance [35], as stress is associated with adipositydespecially visceral obesitydthrough corticosteroid pathways, which are in turn closely related to insulin resistance [36]. Sociocultural differences in how genders are treated and social gender roles may also play a part in this difference [22,24]. The present study also shows that the association between food insecurity and insulin resistance varies by BMI in men. Overweight or obese men had a significant positive association between marginal food security and insulin resistance. The inverse association between insulin resistance and low and/or very low food security in overweight or obese men may be explained by reduced absolute food intake in these states. It is hard to explain the mechanism for this differential association between food security and insulin resistance in both normal weight and

647

overweight or obese men. However, we suppose that insulin resistance may be largely related to either a genetic problem or other external factors such as the low dietary qualify of food insecurity. For overweight or obese men, competing risk could be an issue in which overweight or obese men are at higher risk of progressing to diabetes, regardless of food insecurity. This association could be confirmed or refuted using well designed, longitudinal prospective studies. Normal-weight men in households with very low food security had significantly higher odds of insulin resistance, despite their lower BMI and waist circumference compared with other normalweight men. Public health attention has been focused mostly on the overweight or obese population because obesity is a wellknown metabolic and cardiovascular risk factor [46]. However, normal-weight individuals with insulin resistance that have been designated as having a metabolically healthy normal weight phenotype also face significant health risks [47], especially if they have a higher level of visceral adiposity and a more atherogenic lipid profile, all of which would together contribute to an increased risk of type 2 diabetes and premature coronary heart disease [48e50]. Thus, we suspect that food insecurity might increase the risk of insulin resistance especially in normal-weight men, although it is not associated with weight gain. Previous research has reported that food insecure people are more likely to consume high-fat foods and more juice [47,51]. Hence, diet and eating habits might be different for people with different food security statuses, which might influence the relationship between food insecurity and insulin resistance. Future research will need to evaluate more detailed information on diet and eating behaviors among people within different categories of food insecurity and examine how different diet and eating behaviors affect the relationship between food insecurity and insulin resistance. Our study should be interpreted with consideration of the following limitations: Because the measure of food insecurity in this study was at the household level instead of the individual level, an evaluation of any possible individual-level mechanism(s) between food insecurity and insulin resistance was somewhat limited [52]. Another major limitation might be unmeasured confounding and residual confounding. Moreover, some misclassification of household food insecurity status might have occurred. However, there is no reason to expect that this misclassification would vary by insulin-resistance status. Such nondifferential misclassification of food insecurity would typically reduce the magnitude of the risk estimate between the groups toward the null. In addition, the crosssectional nature of the study design did not allow for evaluation of time sequence or an ability to establish a causal relationship between food insecurity and insulin resistance. Also, the small sample size in the normal-weight women with low food insecurity resulted in a relative SE greater than 30%, which could result in unreliable estimates [27]. Finally, we used HOMA-IR greater than 2.5 to indicate insulin resistance. Although the hyperinsulinemiceuglycemic glucose clamp is considered the gold standard for measuring insulin resistance, this method is too invasive to apply in a large population. HOMA-IR is a simple reliable marker for assessing insulin resistance, showing a high correlation with the hyperinsulinemic-euglycemic clamp technique [32]. In terms of strengths, the present study used a nationally representative sample. Moreover, a well-defined measure of food insecurity and an objective measure of insulin-resistance status provided an opportunity to examine the relationship between food insecurity and insulin resistance, thereby providing further evidence of potential mechanisms responsible for the previously established relationship between food insecurity and insulin resistance. In addition, a relatively large sample size provided adequate power for subgroup analyses with reliable parameter estimates.

648

J. Liu et al. / Annals of Epidemiology 25 (2015) 643e648

In conclusion, food insecurity is associated with insulin resistance in adults without diabetes, and these associations vary by gender and weight status. Future research should evaluate whether improving food security status can reduce the risk for diabetes and cardiovascular disease. References [1] Castillo DC, Ramsey NL, Yu SS, Ricks M, Courville AB, Sumner AE. Inconsistent access to food and cardiometabolic disease: the effect of food insecurity. Curr Cardiovasc Risk Rep 2012;6(3):245e50. [2] Available from: http://www.ers.usda.gov/topics/food-nutrition-assistance/ food-security-in-the-us/key-statistics-graphics.aspx#.UoDyulMZF8E. Accessed February 23, 2015. [3] Pan L, Sherry B, Njai R, Blanck HM. Food insecurity is associated with obesity among US adults in 12 states. J Acad Nutr Diet 2012;112(9):1403e9. [4] Parker ED, Widome R, Nettleton JA, Pereira MA. Food security and metabolic syndrome in U.S. adults and adolescents: findings from the National Health and Nutrition Examination Survey, 1999-2006. Ann Epidemiol 2010;20(5):364e70. [5] Berkowitz SA, Baggett TP, Wexler DJ, Huskey KW, Wee CC. Food insecurity and metabolic control among U.S. adults with diabetes. Diabetes Care 2013;36(10):3093e9. [6] Gucciardi E, Vogt JA, DeMelo M, Stewart DE. Exploration of the relationship between household food insecurity and diabetes in Canada. Diabetes Care 2009;32(12):2218e24. [7] Seligman HK, Laraia BA, Kushel MB. Food insecurity is associated with chronic disease among low-income NHANES participants. J Nutr 2010;140(2):304e10. [8] Drewnowski A. Obesity and the food environment: dietary energy density and diet costs. Am J Prev Med 2004;27(Suppl. 3):154e62. [9] Pereira MA, Kartashov AI, Ebbeling CB, Van Horn L, Slattery ML, Jacobs Jr DR, et al. Fast-food habits, weight gain, and insulin resistance (the CARDIA study): 15-year prospective analysis. Lancet 2005;365(9453):36e42. [10] Hu FB, van Dam RM, Liu S. Diet and risk of type II diabetes: the role of types of fat and carbohydrate. Diabetologia 2001;44(7):805e17. [11] Siri-Tarino PW, Sun Q, Hu FB, Krauss RM. Saturated fat, carbohydrate, and cardiovascular disease. Am J Clin Nutr 2010;91(3):502e9. l M, Kube na A, Macdonald IA. Effects of acute starvation on [12] Duska F, Ande insulin resistance in obese patients with and without type 2 diabetes mellitus. Clin Nutr 2005;24(6):1056e64. [13] Mendoza JA, Drewnowski A, Christakis DA. Dietary energy density is associated with obesity and the metabolic syndrome in US adults. Diabetes Care 2007;30(4):974e9. [14] Seligman HK, Schillinger D. Hunger and socioeconomic disparities in chronic disease. N Engl J Med 2010;363(1):6e9. [15] Eckel RH, Grundy SM, Zimmet PZ. The metabolic syndrome. Lancet 2005;365(9468):1415e28. [16] Rhee EJ, Choi JH, Yoo SH, Bae JC, Kim WJ, Choi ES, et al. The association of unintentional changes in weight, body composition, and homeostasis model assessment index with glycemic progression in non-diabetic healthy subjects. Diabetes Metab J 2011;35(2):138e48. [17] Abbasi F, Brown Jr BW, Lamendola C, McLaughlin T, Reaven GM. Relationship between obesity, insulin resistance, and coronary heart disease risk. J Am Coll Cardiol 2002;40(5):937e43. [18] Choi JY, Ha HS, Kwon HS, Lee SH, Cho HH, Yim HW, et al. Characteristics of metabolically obese, normal-weight women differ by menopause status: the Fourth Korea National Health and Nutrition Examination Survey. Menopause 2013;20(1):85e93. [19] Lee SH, Ha HS, Park YJ, Lee JH, Yim HW, Yoon KH, et al. Identifying metabolically obese but normal-weight (MONW) individuals in a nondiabetic Korean population: the Chungju Metabolic disease Cohort (CMC) study. Clin Endocrinol (Oxf) 2011;75(4):475e81. [20] Karelis AD, Faraj M, Bastard J-P, St-Pierre DH, Brochu M, Prud’homme D, et al. The metabolically healthy but obese individual presents a favorable inflammation profile. J Clin Endocrinol Metab 2005;90(7):4145e50. [21] Ivers LC, Cullen KA. Food insecurity: special considerations for women. Am J Clin Nutr 2011;94(6):1740Se4S. [22] Vlassoff C. Gender differences in determinants and consequences of health and illness. J Health Popul Nutr 2007;25(1):47e61. [23] Svojse EK. An examination of the relationship between levels of food security and depression. Wisconsin: The University of Wisconsin-Milwaukee; 2014. [24] Belachew T, Hadley C, Lindstrom D, Gebremariam A, Michael KW, Getachew Y, et al. Gender differences in food insecurity and morbidity among adolescents in southwest Ethiopia. Pediatrics 2011;127(2):e398e405. [25] Curtin L, Mohadjer L, Dohrmann S, Montaquila J, Kruszan-Moran D, Mirel L, et al. The National Health and Nutrition Examination Survey: sample design, 1999-2006. Vital Health Stat 2 2012;(155):1e39.

[26] Available from: www.cdc.gov/nchs/nhanes/irba98.htm. Accessed August 15, 2014. [27] Available from: http://www.cdc.gov/nchs/nhanes/analytic_guidelines.htm. Accessed August 15, 2014. [28] Available from: http://wwwn.cdc.gov/nchs/nhanes.htm. Accessed August 15, 2014. [29] Resnick HE, Bergman RN, Henderson JA, Nez-Henderson P, Howard BV. Utility of a surrogate measure of insulin resistance in American Indians: the Strong Heart Study. Ethn Dis 2002;12(4):523e9. [30] Bickel G, N M, Price C, Hamilton W, J C, editors. Guide to measuring household food security, Revised 2000. Alexandria, VA: Agriculture USDo; 2000. [31] Matthews D, Hosker J, Rudenski A, Naylor B, Treacher D, Turner R. Homeostasis model assessment: insulin resistance and b-cell function from fasting plasma glucose and insulin concentrations in man. Diabetologia 1985;28(7): 412e9. [32] Bonora E, Targher G, Alberiche M, Bonadonna RC, Saggiani F, Zenere MB, et al. Homeostasis model assessment closely mirrors the glucose clamp technique in the assessment of insulin sensitivity: studies in subjects with various degrees of glucose tolerance and insulin sensitivity. Diabetes Care 2000;23(1): 57e63. [33] Moshfegh AJ, Rhodes DG, Baer DJ, Murayi T, Clemens JC, Rumpler WV, et al. The US Department of Agriculture Automated Multiple-Pass Method reduces bias in the collection of energy intakes. Am J Clin Nutr 2008;88(2): 324e32. [34] Leung CW, Willett WC, Ding EL. Low-income Supplemental Nutrition Assistance Program participation is related to adiposity and metabolic risk factors. Am J Clin Nutr 2012;95(1):17e24. [35] Hadley C, Tegegn A, Tessema F, Cowan J, Asefa M, Galea S. Food insecurity, stressful life events and symptoms of anxiety and depression in east Africa: evidence from the Gilgel Gibe growth and development study. J Epidemiol Community Health 2008;62(11):980e6. [36] Gluck ME, Geliebter A, Lorence M. Cortisol stress response is positively correlated with central obesity in obese women with binge eating disorder (BED) before and after cognitive-behavioral treatment. Ann N Y Acad Sci 2004;1032(1):202e7. [37] Townsend MS, Peerson J, Love B, Achterberg C, Murphy SP. Food insecurity is positively related to overweight in women. J Nutr 2001;131(6): 1738e45. [38] Tayie FA, Zizza CA. Height differences and the associations between food insecurity, percentage body fat and BMI among men and women. Public Health Nutr 2009;12(10):1855e61. [39] Olson CM. Food insecurity in women: a recipe for unhealthy trade-offs. Top Clin Nutr 2005;20(4):321e8. [40] Hanson KL, Sobal J, Frongillo EA. Gender and marital status clarify associations between food insecurity and body weight. J Nutr 2007;137(6):1460e5. [41] Geer EB, Shen W. Gender differences in insulin resistance, body composition, and energy balance. Gend Med 2009;6(Suppl. 1):60e75. [42] Banos G, Guarner V, Perez-Torres I. Sex steroid hormones, cardiovascular diseases and the metabolic syndrome. Cardiovasc Hematol Agents Med Chem 2011;9(3):137e46. [43] Lee DH, Kim YM, Jekal Y, Park S, Kim K-C, Naruse M, et al. Low levels of physical activity are associated with increased metabolic syndrome risk factors in Korean adults. Diabetes Metab J 2013;37(2):132e9. [44] Garaulet M, Perex-Llamas F, Fuente T, Zamora S, Tebar FJ. Anthropometric, computed tomography and fat cell data in an obese population: relationship with insulin, leptin, tumor necrosis factor-alpha, sex hormone-binding globulin and sex hormones. Eur J Endocrinol 2000;143(5):657e66. [45] Haffner SM, Valdez RA, Stern MP, Katz MS. Obesity, body fat distribution and sex hormones in men. Int J Obes Relat Metab Disord 1993;17(11):643e9. [46] Grundy SM. Obesity, metabolic syndrome, and cardiovascular disease. J Clin Endocrinol Metab 2004;89(6):2595e600. [47] Mello JA, Gans KM, Risica PM, Kirtania U, Strolla LO, Fournier L. How is food insecurity associated with dietary behaviors? An analysis with low-income, ethnically diverse participants in a nutrition intervention study. J Am Diet Assoc 2010;110(12):1906e11. [48] Wildman RP, Muntner P, Reynolds K, McGinn AP, Rajpathak S, Wylie-Rosett J, et al. The obese without cardiometabolic risk factor clustering and the normal weight with cardiometabolic risk factor clustering: prevalence and correlates of 2 phenotypes among the US population (NHANES 1999-2004). Arch Intern Med 2008;168(15):1617e24. [49] Florez H, Castillo-Florez S. Beyond the obesity paradox in diabetes: fitness, fatness, and mortality. JAMA 2012;308(6):619e20. [50] Kim MK, Han K, Kwon HS, Song KH, Yim HW, Lee WC, et al. Normal-weight obesity in Korean adults. Clin Endocrinol (Oxf) 2013;80(2):214e20. [51] Leung CW, Epel ES, Ritchie LD, Crawford PB, Laraia BA. Food insecurity is inversely associated with diet quality of lower-income adults. J Acad Nutr Diet 2014;114(12):1943e1953.e2. [52] Available from: http://www.fns.usda.gov/fsec/files/fsguide.pdf.