The literature search was designed to elicit peer-reviewed articles on the relationships between food insecurity (FI) and chronic diseases (CD), and that employed a measure of CD other than obesity. We conducted a systematic search of Medline, JSTOR, Worldcat, Scopus, Public Affairs Index (PAI), Social Services Abstracts, Annual Reviews Online, and Anthro Source electronic databases. In each database, we conducted a series of searches using an algorithm of all possible combinations of keywords in the following two columns:

Column 1                Column 2

Food insecurity        Chronic disease

Food availability       Noncommunicable disease

Food insufficiency      Chronic fatigue

Food shortage          Fibromyalgia

Food sufficiency         Lupus

Hunger                     Obesity

                               Disability

                               Cardiovascular

                               COPD

                              Asthma

                              Hypertension

                              Diabetes

                              Dyslipidemia

These keywords were obtained from prior examination of the literature. This list of chronic diseases is by no means exhaustive, but they were selected to represent a range of diet-related, environmental, and idiopathic conditions while still remaining manageable in scope. Search results were sorted first by relevance, and second, by recency of publication; the first 200 results of each search were screened. No specific publication dates were selected for limiting the literature search, but sorting search results by recency and considering only the first 200 results of each search confined our results to articles published during or after the year 2000. Studies were retained for the literature review only if they met the following inclusion/exclusion criteria:

1) Study includes a CD assessment beyond general health self-report, and a FI assessment beyond receiving government-sponsored nutritional benefits or living in an area designated as a “food desert.”

2) Article reports on the strength of association or the nature of the relationship between FI and CD.

3) Article is a peer-reviewed original study, published in English or with an abstract in English, and deals with humans (not animal models).

4) Reported CD is not skeletal (e.g., arthritis, osteoporosis), not cancer, nor the result of infectious agents (e.g., HIV/AIDS, cervical cancer caused by the HPV virus). Pre-disease risk states, such as dyslipidemia and insulin resistance, were included.

5) Article does not deal with voluntary food restriction, as in cases of anorexia nervosa, or acute hunger, as in famines.

Potentially relevant articles were downloaded and retained for analysis by the second author; this initial search process resulted in 196 potential articles for inclusion, which were then reviewed by the first author and reduced to 26 articles based on the inclusion and exclusion criteria listed above. To ensure thoroughness, we subsequently performed a cited-reference search (“reverse search”) for each article using Google Scholar and reviewed the first 30 results of each search. This resulted in an additional 23 articles fitting the inclusion criteria. Finally, we mined the bibliographies of the retained articles, which led us to an additional 2 meeting the inclusion criteria.

In the qualitative and case-control literature, decisions about the inclusion or exclusion of studies were not always straightforward because assessment of FI or CD was often not quantified or was used as a sample selection criterion rather than a study variable. In such cases, we skimmed results and discussion sections and included them when they involved explicit discussion of FI-CD relationships.

The two authors independently reviewed each article and extracted the following data: study location, design (sample and study type), measure of food insecurity, measure of chronic disease, reported association between the two, control variables used in multivariate analyses (if any), and authors’ discussions about the nature of the observed association, or lack thereof. In cases where there were discrepancies in the data extracted from the articles, the first author re-reviewed the article.

The 15 studies included in the literature review that controlled for BMI in their analyses consistently reported associations between FI and CD risk independent of obesity. Seligman et al. [48], for instance, found that in a nationally representative sample of Americans, food insecurity was independently associated with diabetes. Although obesity among women was also associated with mild food insecurity, obesity accounted for only 20% of the increased odds of diabetes. Importantly, these studies indicate that a significant portion of the FI-CD relationship cannot be explained by obesity. We therefore urge researchers to include direct measures of CD risk biomarkers (such as high-sensitivity C-reactive protein assay for cardiometabolic risk, or lung function for asthma severity) along with BMI. Moreover, we underscore the need for biopsychosocial studies integrating qualitative work, biomarkers for CD and risk status, and standardized measures of FI to identify some of the apparent sources of association between FI and CD independent of obesity. This appears to be important both for diet-related and non-diet-related CD; among the studies demonstrating relationships between CD and food insecurity independent of obesity were both those focusing on diet-related diseases such as type 2 diabetes, and those focusing on non-diet-related diseases such as asthma.

These 15 studies are in the minority; little of the FI-CD research is designed to include statistical tests of the presence or absence of a mediating role for obesity in the FI-CD association. To correct for this, we recommend that even if researchers adopt more direct measures of cardiometabolic risk in their studies, they still include BMI as a control variable to assess the portion of the FI-CD relationship not accounted for by BMI.

Many studies documented a FI-CD relationship without speculating about its causes, and those that did so almost universally addressed causes as theoretical possibilities rather than performing statistical tests to identify potential mediators. As a whole, the studies suggested three main categories of mediators. Each is summarized below.

The first, and most common, pathway that researchers proposed to link FI and CD is a broad category that we refer to as “economic mechanisms”. In households with economic constraints, FI is more likely, as are a suite of other stressors that may interact cyclically with health to produce CD risk even when obesity is not present. As Crews et al. [38] and Tarasuk et al. [51] noted, unemployment related to chronic illness disability may be an important economic driver the CD-FI association. In such cases, the high cost of CD medications may leave insufficient money for nutritious food, and a lack of nutritious food exacerbates many CDs even in the absence of obesity, leading to higher treatment costs, medication nonadherence, and increased FI, in a cyclical relationship [36] [37] . Some of the work included here demonstrates that poverty leads to both FI and reduced healthcare usage (including medication nonadherence) [52] [53] .

The six studies that included qualitative data also underscored the relevance of economic mechanisms linking FI and CDs. These studies demonstrated that low-income people with type 2 diabetes struggled to adhere to diabetic diets because of economic constraints including limited household finances, geographic unavailability of healthy foods, and lack of control over their own diets, rather than because of lack of awareness about the importance of dietary control [54] [55] [56] . These studies particularly highlighted the diabetes-inappropriate food available at food banks and the fact that dependence on food aid constrains one’s ability to choose what to eat [32] .

The wide range of pathways falling under the economic explanation attests to a social reality: that food insecurity rarely occurs in isolation; rather, it is only one of many physical, mental, and social health problems that households face when dealing with economic constraints. The studies that assessed psychological, social, and health outcomes found that FI is only one of many variables associated with CD risk among marginalized groups; other important CD predictors were depression and poor social support; culturally specific distress, gastrointestinal infection, low maternal education, and child learning disorders; and homelessness, stressful life events, low birthweight, and parental anxiety [34] [57] [58] . Food insecurity is both an index and an engine of poverty, and poverty is one of the most systematically health-harming forces to which humans can be exposed [59] .

A second pathway potentially linking FI and CD in the absence of obesity is cyclically dysregulated eating patterns among FI individuals, such as reduced intake during food shortage followed by overconsumption when food is available, which leads to metabolic disruption that can lead to metabolic diseases even when a person is not obese [28] [40] [48] [60] . The eating disorder literature has long demonstrated that cyclical food restriction and bingeing interferes with normal metabolism via dysregulation of the HPA axis, which in turn appears to be related to visceral fat deposition (a strong risk factor for many chronic diseases) [61] [62] [63] . There is a parallel body of literature suggesting direct relationships between FI and dysregulated eating patterns [64] [65] . Food assistance programs might play a role in dysregulated eating because they emphasize processed foods and typically disburse benefits once monthly, thus leading to monthly cycles of feast-and-famine [1] [4] [66] [67] . Dysregulated eating behaviors―potentially perpetuated by reliance on once-monthly food aid or food banks―may predispose individuals to chronic diseases via metabolic dysregulation without the necessary precondition of obesity.

A third possible pathway linking FI and CD in the absence of obesity is high exposure to environmental toxins among people suffering from FI. Sattler and Bhargava [33], for instance, note the greater likelihood of smoking in low-income households as a potential risk factor for CDs independent of obesity. Along the same lines, Shiue [68] points out that people suffering from FI are also more likely to be exposed to high levels of environmental contaminants known to induce chronic diseases, such as bisphenol-a (BPA; an endocrine disrupter), pesticides (also endocrine disrupters), and antimony (a chemical element linked to liver dysfunction).

Strikingly few of the studies included in this review discussed biosocial pathways that might connect FI and CDs, such as the psychoneuroimmunological connections between distress and chronic health problems. Mangini et al. [45] make a nod to the whole-person approach to understanding FI-CD risk, pointing out that the associations between household poverty and child asthma must be understood within the framework of a life course model of health disparities. Pérez-Escamilla et al. [47] similarly point out that the chronic stress of FI could lead to higher stress hormones such as cortisol, which promotes abdominal fat storage, leading to type 2 diabetes and hypertension. None of these studies, however, considers the possibility that stress activation could lead to these CDs independently of the mechanism of abdominal fat storage-for instance, through generalized inflammation, which has been shown to be a strong predictor of chronic disease even when controlling for socioeconomic and dietary variables [69] .

The literature was generally biased toward cross-sectional, quantitative studies based in North America. More longitudinal and qualitative studies are needed, as are studies conducted in low- and middle-income countries. Several studies emphasized the unique challenges faced by women, but work is needed on the qualitative and quantitative associations between FI and CD in men specifically, since men may experience FI and CD differently than women both physiologically and psychologically. Moreover, because studies addressing asthma and dyslipidemia were, as a whole, inconclusive about their relationships with FI, more work is needed on these potential relationships. No studies addressing chronic fatigue syndrome, fibromyalgia, lupus, or COPD were found during the review process. This may seem logical, given the widespread assumption that non-lifestyle-related diseases are less likely to be affected by food insecurity than are those directly related to diet or lifestyle. However, studies demonstrating clear links between food insecurity, stress response, and body inflammation suggest, at very least, that there is a potential link between non-diet-related diseases and food insecurity [69] . This warrants further exploration.

Further study is also needed to determine if there are differences in the FI-CD relationship in low- and middle-income settings as opposed to industrialized countries; the three studies in Malaysia documented no association between FI and CD, suggesting that this relationship may work differently in less-developed economies. Likewise, further work on the role of gender, minority race/ethnicity, and age with relevant comparison groups would help elucidate how these demographic characteristics may shape the FI-CD relationship.

Finally, the large proportion of studies publishing positive results could be of concern. While it is entirely plausible that this reflects a secular association between CD and FI, it could also reflect publication bias toward positive study results. We therefore encourage researchers who find no positive association between CD and FI to seek publication.

This review was not designed to include the literature on the associations between health behaviors and food insecurity, which is substantial. The review also did not include studies conducted among populations with HIV/AIDS, which under treatment with regular antiretroviral drugs becomes very much like a chronic disease. There is significant literature, however, exploring relationships between FI and HIV/AIDS.