Direct and Indirect Pathways Between Food Insecurity and Glycemic Control in Adults with Diabetes.

INTRODUCTION

As a leading cause of death and disability worldwide, diabetes is a major public health concern.1 In the USA, 14.7% of all adults have diabetes, an increase from just over 10% in 2001–2004.1 Although lowering blood glucose levels, blood pressure, and low-density lipoprotein cholesterol can lead to better outcomes, only 11.1% of adults with diabetes meet the recommended clinical criteria for managing risk factors.1 This results in increased rates of complications and premature mortality for many living with diabetes.1

Extensive research identifies social determinants of health as a leading driver of poor health outcomes in adults with diabetes.2–5 Food insecurity, commonly defined as not having access to enough food for an active, healthy life, is a social determinant shown to influence the prevalence of diabetes, diabetes self-management, diabetes complications, glycemic control, quality of life, and healthcare expenditures in adults with diabetes.6–14 Multiple pathways have been hypothesized through which food insecurity influences health outcomes.8 These include psychological, behavioral, and biological pathways, including the potential for multiple mechanisms to exist simultaneously.8

One pathway highlighted by a recent state of the science conference involves food insecurity triggering the stress pathway and resulting in activation of the hypothalamic pituitary adrenal axis and an increase in visceral fat accumulation.8 Food insecurity is associated with higher body mass index (BMI) and a higher risk of being overweight; however, this relationship appears to differ by gender being most often noted in women.15–17 A study testing this pathway using structural equation modeling found that food insecurity was indirectly associated with higher BMI through greater distress, though this relationship did not hold for diet quality.18 Though much of the work in food insecurity has focused on relationships with BMI, other anthropometric and body composition measures can be used to assess overall health and an individuals’ nutritional state.19

Poor diet associated with food insecurity is a second major pathway that is hypothesized to increase inflammation, impacting immune functioning and ultimately worsen cardiometabolic health.8 Studies show differences in diet for food secure and food insecure individuals, including food insecure adults reporting higher carbohydrates, less protein, and less total fat. (Lee 2018) While diet quality in the USA for adults with diabetes has risen in the past decade, differences by food insecurity status remained the same.21 Using the Healthy Eating Index (HEI) as a measure of diet quality, one study found that individuals who were both food insecure and had poor diet quality had increased odds of higher HbA1c.22 However, a second study using the HEI in adults with diagnosed diabetes found that food insecurity was associated with increased HbA1c, regardless of whether individuals reported a high-quality or low-quality diet, suggesting diet alone may not explain the relationship.23

Though multiple pathways have been hypothesized, few have been tested empirically. Identifying mechanisms of influence is critical for developing effective interventions that address the influence of food insecurity on diabetes outcomes. Therefore, the objective of this analysis was to investigate the direct and indirect relationship between food insecurity and glycemic control focusing on three possible pathways: body composition, anthropometrics, and healthy eating.

METHODS

The National Health and Nutrition Examination Survey (NHANES) is an all-encompassing initiative carried out by the National Center for Health Statistics (NCHS), a division under the Centers for Disease Control and Prevention (CDC) in the USA. This program is designed to evaluate the health and nutritional well-being of the entire US population, serving as a comprehensive repository of nationally representative health data.24 The NHANES survey occurs biennially, with each cycle comprising a random sample of 5000 individuals from 15 different regions throughout the USA annually. Each participant serves as a representative sample for approximately 65,000 other individuals akin to them.24 The survey encompasses five primary components, namely demographic data, dietary information, physical examinations, laboratory analyses, and questionnaire responses.24

This analysis was conducted with data from adults 18 years old and above, who answered food security questions and indicated yes for whether a health professional had ever told them they had diabetes (N = 3055). IRB review and ethics approval was not required as data was publicly available and non-identifiable.

Blood samples were collected and analyzed for various health markers, including HbA1c levels. HbA1c is a measure of average blood glucose over a 3-month period and is commonly used in the diagnosis and management of diabetes. The continuous HbA1c was used an outcome variable.

The questionnaire on food security gathered data regarding household food security, individual food security, and participation in food programs such as food stamp authorization and WIC. Households with an income four times above the U.S. Department of Health and Human Services (DHHS) poverty guidelines did not undergo the U.S. Household Food Security Survey Module (FSSM) questions. Participants in these households were assigned a code of “6 (question not administered).” For analytical purposes, the adult food security variable was utilized. This variable was categorized, with adult high food security and marginal food security considered food secure, and low food security and very low food security categorized as food insecure.24

Dual-energy x-ray absorptiometry (DXA) was used to scan the whole skeleton to access the changes that happened due to change in age, gender, and race by measuring bone mineral content (BMC), bone mineral density, total body fat, and lean mass measurements.24 Body composition components include total fat and total lean. NHANES then calculates the fat mass index (FMI) by dividing the total fat in kilograms by height in meters square and calculates the lean mass index by diving total lean in kilograms by height in meter square. Each variable was incorporated into the analyses in a continuous format.

In NHANES, anthropometric measurements were collected in a mobile examination center (MEC) to assess growth and distribution of body fat. The analysis used two anthropometric measurements: body mass index (BMI) and waist circumference. BMI was calculated by dividing weight in kilograms by the square of height in meters (kg/m2). Both height and weight were measured in MEC and BMI was analyzed as a continuous scale. Waist circumference was measured in the MEC while the participant was standing by marking a horizontal line above the right iliac and crossing it to indicate the midaxillary line. The measurement of the waist circumference was then completed by using a measurement tape at the mark and normal expiration.24 The waist measurement was measured in centimeters and analyzed at a continuous scale.

Computation of the Healthy Eating Index-2015 (HEI-2015) involved utilizing data collected from the food questionnaire in NHANES, with the calculation specifically relying on 24-h dietary recall information.24,25 The Healthy Eating Index comprises 13 components, including total vegetables, greens and beans, total fruit, whole fruit, whole grains, total dairy, total protein, seafood and plant protein, fatty acids, sodium, refined grain, saturated fats, and added sugar. Out of 13 components, three components—Whole Grain, Total Protein, and Fruit Vegetables—were selected for analysis in addition to the overall Healthy Eating Index score in this analysis based on strength of initial correlations with both food insecurity and HbA1c. All variables were incorporated into the analyses in a continuous format.

Covariates included age (continuous), sex (male, female), education (< high school, high school/GED, some college, college graduate), race/ethnicity (non-Hispanic Black, non-Hispanic White, Hispanic, and Other race/ethnicity), income (less than 130% poverty level, 130–185%, and greater than 185% poverty level), and marital status (never married, married, separated/divorced, widowed).

Structural equation modeling (SEM) was used to investigate the direct and indirect relationships hypothesized in Fig. 1 to exist between food insecurity and glycemic control. Based on available variables in the NHANES dataset, possible mediators were identified for each of the three hypothesized pathways. Body composition variables included total fat, total lean, fat mass index, and lean mass index. Anthropomorphics included body mass index (BMI) and waist circumference. Healthy eating included sub-scales of the healthy eating index (HEI) including whole grain, total protein, and fruits and vegetables, as well as the overall HEI score. Initial correlations were run between food insecurity, each possible mediator, and HbA1c to identify the strongest relationships for investigation in SEM models. Appendix Table 1 provides correlation coefficients for all relationships between food insecurity, glycemic control, body composition, anthropomorphics, and healthy eating index variables. Total lean was identified as the strongest correlation for body composition, waist circumference was identified as the strongest correlation for anthropomorphics, and overall HEI score was identified as the strongest correlation for healthy eating.

Using Stata v17, SEM was performed using the “sem” command and the maximum likelihood estimation procedure.26–28 The “mlmv option” was used to retain variables in the analysis instead of dropping individuals from the analysis due to listwise deletion. First, a full model was run (see Appendix Table 2), followed by a trimmed model, removing paths shown to be not significant in the full model (indicated by grey lines in Fig. 2). Direct, indirect, and total effects were estimated and the model was adjusted for age, sex, race/ethnicity, income, education, and marital status based on theoretically relevant confounders. Structural relationships were evaluated through direction and magnitude of path coefficients. In addition, model fit was evaluated using multiple fit statistics, per best practice standards. Root square mean error of approximation (RSMEA) identifies good fit if less than 0.05, comparative fit index (CFI) identifies good fit if greater than 0.9, and Tucker-Lewis fit index (TLI) identifies good fit if greater than 0.9.29 Significance of results was assessed using a p-value < 0.05.

RESULTS

Table 1 provides sample characteristics for the 3055 individuals in NHANES 2011–2018 who reported diabetes. The mean age was 61.9 years and 47.8% of the sample were women. Twenty-five percent of the sample reported food insecurity. Individuals who were food insecure were more likely to be younger, be female, have less than a high school education, not be insured, have a higher BMI, and have higher HbA1c.

Figure 1 shows the hypothesized model, Fig. 2 shows the final trimmed model with standardized coefficients, and Table 2 shows direct, indirect, and total effects of the final trimmed model. Standardized beta estimates are interpreted as an estimate increase in the outcome for every 1 standard deviation change in the predictor. Significant indirect effects indicate pathways through which variables influence outcomes. The final model was well fit (χ2(24) = 697.72, p < 0.001, RMSEA = 0.00, CFI = 1.0, TLI = 1.0, pclose 1.0) and after adjustment for age, sex, race/ethnicity, income, education, and marital status showed higher food insecurity was directly associated with higher HbA1c (B = 0.06, p < 0.001) and indirectly associated through waist circumference and HEI score (B = 0.005, p = 0.002) in adults with diabetes. Food insecurity was significantly directly associated with higher waist circumference (B = 0.06, p < 0.001) and lower healthy eating (B = −0.04, p = 0.01). Higher waist circumference (B = 0.05, p = 0.002) and lower healthy eating (B = − 0.04, p = 0.007) were significantly directly associated with higher HbA1c. Body composition was not a significant pathway. Therefore, food insecurity was associated with glycemic control both directly and indirectly via the pathways of waist circumference and healthy eating.

DISCUSSION

In a nationally representative sample of adults with diabetes, 25% reported food insecurity, and higher food insecurity was directly associated with higher HbA1c (worse glycemic control). Higher food insecurity was also indirectly associated with HbA1c via higher waist circumference and lower healthy eating index (worse dietary quality). Finally, food insecurity was significantly associated with higher waist circumference and lower healthy eating index. In this study testing three possible pathways, body composition was not found to be a significant pathway through which food insecurity was associated with glycemic control.

This study adds unique findings to the literature by identifying two promising pathways to target interventions aimed at addressing the influence of food insecurity on adults with diabetes. First, we found that waist circumference and overall healthy eating were significant pathways, suggesting interventions should incorporate evidence-based components that support weight loss and healthy eating. Existing evidence supports the relationship between waist circumference and food insecurity among adults30,31; however, little has been done to develop targeted interventions, specifically among adults with diabetes experiencing food insecurity. The ADA outlined key recommendations for addressing obesity and weight management among adults with diabetes, including the need to account for social risk factors such as food insecurity.32 However, current interventions focus largely on the individual level with limited attention placed on structural factors underlying inequities in food access, nutrition, and physical activity options. Promoting food security, access to healthy food options, and weight management simultaneously are essential for addressing pathways identified in this study. Therefore, identification of social and public policies that can mediate pathways through which individuals have differential access to healthy food and weight management options is critical.33

Second, the majority of work conducted on weight and food insecurity has focused on BMI.15–18 This analysis found that waist circumference may be a more relevant measure of influence on which to focus. Evidence shows that waist circumference is a more precise measure of fat distribution, compared to BMI, and is strongly associated with disease risk as well as mortality.34–37 A study of the relationship between BMI and waist circumference on cardiovascular outcomes noted that optimal thresholds differed by racial/ethnic group for both BMI and waist circumference.38 Racial/ethnic groups at highest risk by different thresholds also differed highlighting the importance of incorporating additional anthropomorphic measures into studies and considering differences by gender and racial/ethnic representation.38 Of note, our findings indicated that body composition measures were not a significant pathway through which food insecurity influences glycemic control in adults with diabetes. Given the time and resource-intensive process required to capture body composition measures, this suggests interventions focused on addressing the influence of food insecurity on glycemic control can focus instead on anthropomorphic measures that are easier to capture.

A growing body of work exists testing interventions targeting the influence of food insecurity on health. A recent review of Food Is Medicine interventions, including medically tailored meals/groceries and produce prescriptions/vouchers, found significant improvements in HbA1c, healthy eating, depressive symptoms, diabetes distress, and quality of life.39 However, Food Is Medicine interventions for adults with diabetes showed limited effects on blood pressure and BMI, which authors suggested indicated a need for more intensive health education and behavior change support.39 Most studies used one-group designs or had small sample sizes and reported large attrition rate,39 so results should be considered cautiously, but do suggest that tailored interventions focused on addressing food insecurity in adults with diabetes may improve diabetes outcomes. This study can inform ongoing work and future adaptations of interventions by highlighting the importance of incorporating multiple components within intervention designs and specifically focusing on weight loss and healthy eating.

Strengths of this study include using a national dataset and methodology to allow investigation of multiple pathways for the influence of food insecurity on glycemic control in adults with diabetes. However, there are some limitations worth noting. First, data is cross-sectional and cannot speak to causality. Second, while this study focused on three possible pathways, additional pathways may exist that need further investigation. Finally, as the goal of this study was not to investigate differences in pathways between groups of individuals, we cannot comment on if the strength or importance of any of the three pathways investigated differ by demographic factors such as age, race, or gender. Future work should investigate possible differences to help in targeting interventions.

In conclusion, in a study using US national data, waist circumference and overall healthy eating index are significant pathways through which food insecurity is associated with worse glycemic control in adults with diabetes. Based on these results, future interventions targeting adults with diabetes that are experiencing food insecurity should incorporating multiple components within the design and target weight loss and healthy eating.