A Scoping Review of Nutritional Biomarkers Associated with Food Security

Food insecurity affects more than 40 million individuals in the United States and is linked to negative health outcomes due, in part, to poor dietary quality. Despite the emergence of metabolomics as a modality to objectively characterize nutritional biomarkers, it is unclear whether food security is associated with any biomarkers of dietary quality. This scoping review aims to summarize studies that examined associations between nutritional biomarkers and food security, as well as studies that investigated metabolomic differences between people with and without food insecurity. PubMed, Embase, Scopus, and AGRICOLA were searched through August 2022 for studies describing food insecurity and metabolic markers in blood, urine, plasma, hair, or nails. The 78 studies included consisted of targeted assays quantifying lipids, dietary nutrients, heavy metals, and environmental xenobiotics as biochemical features associated with food insecurity. Among those biomarkers which were quantified in at least five studies, none showed a consistent association with food insecurity. Although three biomarkers of dietary quality have been assessed between food-insecure versus food-secure populations, no studies have utilized untargeted metabolomics to characterize patterns of small molecules that distinguish between these two populations. Further studies are needed to characterize the dietary quality profiles of individuals with and without food insecurity.


Introduction
Food insecurity (FI) is a social determinant of health that affects over 40 million individuals in the United States [1]. Worldwide, the prevalence of moderate to severe food insecurity is approximately 22% [2]. Food insecurity is defined as a lack of reliable access to sufficient quantities of nutritious food and can be classified at an individual or household level into high, marginal, low, and very low food security [1]. It is associated with negative health outcomes such as obesity [3], cardiovascular disease [4], asthma [1], exacerbation of type 2 diabetes [5], and overall mortality [6]. The effects of food insecurity span the spectrum of age, associating with increased hospitalizations in childhood [7] and functional limitations in the elderly [8]. The COVID-19 pandemic has further exposed existing disparities in food security, highlighting how race, income, and mental health characteristics affect vulnerability to FI [9].
For decades, policies have concentrated on combatting food insecurity by ensuring adequate caloric supply [10]. However, many of the present-day adverse health effects of food insecurity are attributed to reduced dietary quality, not quantity [11]. Organizations such as the USDA are beginning to place a new emphasis on nutrition security, distinct relationship. Additionally, if only a defined subset of participants demonstrated a significant relationship between food security status and a biomarker, the finding was counted as showing a significant relationship for that population.
Due to heterogeneity between populations, geographies, food security measures, age groups, cultural practices, adjusted covariates, and collection methods, weighted quantitative analysis was not performed. Instead, the directions of relationships between reviewed biomarkers and food insecurity were tabulated for each study and results were tallied into categories showing positive, negative, or non-statistically significant (e.g., p > 0.05) association.

Results
A total of 6850 studies were identified by the search query. Following removal of 4090 duplicates, 2760 studies were screened on the basis of titles and abstracts, leaving 274 for full-text review. Of these, 78 met inclusion criteria and were included in the present analysis. Figure 1 shows the PRISMA diagram for the screening process.

Study Characteristics
The key characteristics of the 78 included studies are summarized in Table A1  . All included studies were published between 2000 and 2022. Thirty-two studies (41%) were conducted in the US; the remainder were conducted in countries including Malaysia, Iran, Azerbaijan, Colombia, India, Mexico, Bangladesh, Ethiopia, Kenya, Vietnam,

Study Characteristics
The key characteristics of the 78 included studies are summarized in Table A1  . All included studies were published between 2000 and 2022. Thirty-two studies (41%) were conducted in the US; the remainder were conducted in countries including Malaysia, Iran, Azerbaijan, Colombia, India, Mexico, Bangladesh, Ethiopia, Kenya, Vietnam, Canada, Brazil, Ecuador, Uganda, Ghana, Tanzania, Mongolia, and Pakistan. Two studies (2.6%) used a longitudinal cohort study design, while the remaining seventy-six (97.4%) were cross-sectional. Nineteen studies (24.4%) used NHANES data on food security and serum micronutrient levels. The USDA food security scales, including the HFSSM (which has validated translations and a shortened 6-item version) and the Adult Food Security Survey Module, were used by 35 (44.9%) of the studies. The next most commonly used tool was the Household Food Insecurity Access Scale (HFIAS), which was used by 15 (19.2%) of the studies. Among the pediatric studies, food security was measured at the household level in 25 out of 27 (92.6%) studies, while, among adult studies, household measures were used in 42 out of 47 (89.4%) studies (the remaining studies assessed food security through derivative measures). Nearly all (73 out of 78) of the studies obtained biomarker data solely from blood samples; three used urine samples, and two used both blood and urine.

Associations between Food Insecurity and Biomarkers
Across the 78 included studies, 59 unique biomarkers were analyzed for their association with food insecurity. The total numbers of studies showing statistically significant positive or negative associations, as well as absence of association, are depicted for each biomarker in Figure 2. The ten biomarkers investigated by at least five studies each were total cholesterol, high-density lipoprotein (HDL), low-density lipoprotein (LDL), triglycerides, ferritin, folate, zinc, vitamin A, vitamin B12, and vitamin D. No uniform associations were found between food security and any of these biomarkers. Among the biomarkers evaluated by between two and five studies, vitamin E, alpha-carotene, and bisphenol A (BPA) showed exclusively unidirectional associations across studies (BPA was positively associated with food insecurity, while vitamin E and alpha-carotene were negatively associated). All of the remaining biomarkers were evaluated by either only one study, or did not demonstrate consensus of relationship directionality. Beta-cryptoxanthin, alphacarotene, and beta-carotene were the only identified biomarkers known to be associated with dietary quality.
Subgroup analyses were performed by geographic location (US vs. non-US) and age-group (non-pregnant adult vs. child under 18). No uniform directional relationships between food insecurity and any biomarker were found for any of these subpopulations (supplemental Figures S2-S6). Similarly, subgroup analysis limited to only NHANES data did not show any new or notably strengthened relationships (supplemental Figure S7).  Counts of research articles demonstrating positive, negative, or null associations between levels of a given biomarker and the presence of food insecurity. Certain biomarkers investigated by only one study were grouped together under broader categories, defined below. The sign following each biomarker indicates its individual relationship with food insecurity, with (+) indicating the biomarker is positively associated with the presence of food insecurity, (0) indicating the biomarker is not associated with the presence of food insecurity, and (-) indicating the biomarker is negatively associated with the presence of food insecurity.

Discussion
The objective of this scoping review was to summarize the existing literature investigating the association between food security status and physiologic concentrations of nutritional biomarkers. No studies were found that used untargeted metabolomics to investigate this question. Among the studies using targeted analyses, the 10 biomarkers investigated by at least five studies-total cholesterol, HDL, LDL, triglycerides, ferritin, folate, zinc, vitamin A, vitamin D, and vitamin B12-all showed predominantly no relationship with food security status.
Previous reviews have examined the relationships between food insecurity and intake levels of specific nutrients. There is no or very limited evidence suggesting adverse association between food insecurity and intake of vitamin B12, folate, or iron [99]. Likewise, there is very limited evidence of association between food insecurity and total fat intake [99,100]. More evidence exists for adverse association between food insecurity and intakes of vitamin A and zinc [99]. While intake levels serve as an important proxy for serum concentrations, metabolic and environmental differences make it impossible for them to completely reflect physiologic concentrations.
A notable finding of the present review was the relative paucity of studies evaluating patients with and without FI for biomarkers of specific food categories associated with dietary quality; none of the 10 most-identified biomarkers were dietary metabolites. Total cholesterol, LDL, HDL, triglycerides, and ferritin are susceptibility biomarkers, which can measure the impact of a nutritional exposure on host physiology, while folate, iron, vitamin A, vitamin D, and vitamin B12 are micronutrients (non-specific exposure biomarkers) [101]. Only 3 of the 59 identified biomarkers were dietary metabolites associated with specific food group exposures (per the collection identified in [17])-beta-cryptoxanthin, alpha-carotene, and beta-carotene. These showed predominantly negative associations with FI, but were reviewed by fewer than five studies each. Thus, the biomarkers for which there were the most data were biomarkers of general nutritional status, not specific dietary exposure.
The lack of consistent directional relationships between food security status and physiologic concentrations of biomarkers may have had to do with the inherent limitation of using food insecurity as a proxy for nutrition quality. Individuals with food insecurity can be food-insecure in different ways, most notably by over-versus undernutrition. These states would be expected to have different biochemical profiles, and combining them under the umbrella of FI increases the heterogeneity of the population, potentially blurring relationships. Nutrition insecurity, on the other hand, would likely have a more consistent biochemical signature, but, at present, is not as robustly measured in population-level studies as FI. Both measures may nevertheless share explanations for relationships with physiologic concentrations of specific biomarkers. To this aim, a summary of explanations extracted from included studies for the relationships, or lack thereof, between the 10 mostreviewed biomarkers and food security is provided below.

Lipid-Related Markers-Total Cholesterol, HDL, LDL, and Triglycerides
Overall, out of the 78 total studies in this review, the majority of those examining lipid markers (total cholesterol, HDL, LDL, and triglycerides) found no significant association with food insecurity. Generally, researchers cited sample size, non-random sampling, non-fasting serum measurements, and unexpectedly high prevalence of dyslipidemias as explanations for this lack of relationship [52,96]. Nevertheless, the predominant lack of association between food insecurity and lipid markers has also been identified elsewhere, as noted in a recent meta-analysis [102]. Similar to the present findings, the authors of the meta-analysis also concluded that current data do not suggest that there are consistent relationships between dyslipidemias and food insecurity status.
Of the studies that identified significant associations between the presence of dyslipidemias and food insecurity, many cited obesity as a mediating factor. Obesity is known to play a key role in generating many of the hallmarks of dyslipidemia, including low HDL-C. Increased remnants of chylomicrons and very-low-density lipoprotein (VLDL) molecules paired with reduced lipolysis in obesity result in diminished HDL as well as diminished HDL function [103]. Moreover, three out of the eight studies that found a positive association between FI and triglyceride levels were conducted with women [49,77,92]. Studies have reported that women, but not men, who experience food insecurity are more likely to have overweight or obesity [104,105]. Therefore, the high prevalence of hypertriglyceridemia was hypothesized to be mediated by obesity and differences in abdominal fat distribution among these women [73,77,106]. Sex-based differences in hormonal regulation and metabolism of lipids may also explain some of this association [73].
Other explanations for associations between FI and dyslipidemias derive from numerous factors. One hypothesis is centered on individuals with FI demonstrating decreased consumption of antioxidants. Decreased antioxidant defense reduces protection against the peroxidation of HDL, and oxidatively modified HDL demonstrates loss of its antiatherogenic properties [92,107]. Other hypotheses posit that individuals with FI may participate in limited physical activity and experience greater psychosocial stress, which may explain associations with dyslipidemia due to dysregulated cortisol and metabolic hormones [73]. Finally, one author indicated that food-insecure individuals may display patterns of overeating during times of food availability in expectation of future food shortages [49]. If this behavior occurred during the time period of sample collection, it could have explained the lipid marker elevation.

Zinc
Zinc was investigated as a biomarker in eight studies, in which six found no association with food insecurity [57,64,65,67,80,83] and two found a negative association [44,81]. Of the two studies that found an association, one was conducted on Colombian children 1-4 years of age [81] and one was conducted on pregnant women in Ethiopia [44]. Decreased access to zinc-rich foods such as meats, as well as a higher prevalence of plant-based diets rich in zincabsorption inhibitors such as fiber and phytic acid, were proposed as explanations for why these food-insecure populations were more likely to have lower zinc concentrations. It was also noted that zinc serum concentrations can be altered by contamination when acquiring or processing a sample [81]. This collection technique-associated variance is important to consider in the interpretation of results from large-scale studies. The six studies that found no association studied US children [57], Bangladeshi children living in a slum [65], Ghanian women aged 18-35 [83], Brazilian preschoolers [80], and Colombian children [64,67]. In some of these countries, culturally relevant dietary sources of zinc are generally uncommon, so the whole population may have low zinc levels [80]. This would lower observed differences between food-secure and -insecure populations. Likewise, in populations with broad access to affordable foods rich in zinc, differences in zinc levels would once again be minimized.

Ferritin
A total of 27 studies analyzed the relationship between ferritin concentration and food insecurity, with 7 finding a negative association, 1 finding a positive association, and 19 finding no association [23][24][25]31,[34][35][36]42,47,48,[53][54][55][56][57][58]63,65,67,72,76,78,79,82,83,85,94]. The studies that identified a negative association attributed this relationship to chronic iron deficiency in the diet [54], parents giving up iron-rich foods to their children and not getting enough iron themselves [54], increased inflammation reducing iron absorption [55], decreased access to heme sources of iron [76], decreased knowledge about iron supplementation [76], shorter inter-pregnancy interval with insufficient time to restore iron stores [48], and increased rates of parasitic infections that may reduce iron absorption [47]. The single study that found a positive association was conducted among Inuit adults who had high consumption of traditional foods, which are often rich in iron [108]. Moreover, the authors noted that traditional food consumption varies throughout the year, so, if the data were collected during a different season, the result may have differed. Among the studies that found no association, explanations for the results included high background of iron deficiency among the population [24,58], effective distribution of iron-rich foods through food assistance programs [25], improved iron absorption from decreased inflammation (achieved through high intakes of traditional foods rich in anti-inflammatory long-chain polyunsaturated fatty acids) [53], limited iron absorption inhibitors in traditional foods [53], infection increasing serum ferritin level [63], high levels of iron fortification in common foods [57], and high levels of iron supplementation [85].

Vitamin A
Of the 15 studies that examined the relationship between food insecurity and vitamin A [22,24,26,31,38,46,56,63,65,67,79,80,83,94,98], 7 found that food insecurity was significantly associated with lower serum vitamin A levels [24,26,31,56,79,83,94], and the remaining 8 studies found no significant association [22,38,46,63,65,67,80,98]. The studies which found lower vitamin A levels in populations with FI cited dietary deficiency, lack of sanitation, social and economic deprivation, and serious infections that deplete vitamin A stores as factors explaining the relationship [24,26,94]. Fourteen of the sixteen studies took place outside of the United States, namely, in South Asia, South America, and Africa. It is well-known that vitamin A deficiency clusters in regions of poverty, limited infrastructure, and high prevalence of infectious disease [109]. However, it is notable that nine studies found no significant associations. Reasons for lack of association include low prevalence of severe food insecurity due to sampling methods, small sample size, well-established food fortification programs, and robust hepatic maintenance of stable serum vitamin A levels [22,46,63,110].

Folate
Of the eight studies that examined the relationship between food insecurity and serum folate [31,33,45,57,67,79,84,89], three found that FI was significantly associated with lower serum folate levels [33,45,84], and the remaining five studies found no significant associations [31,45,57,67,79,89]. Studies that found lower folate levels cited severe food insecurity, poverty, decreased mobility, and decreased ability to care for oneself, particularly among the elderly, as factors contributing to the negative association [33,84]. In studies that found no significant associations between food insecurity and serum folate levels, government fortification programs and parents protecting children from food insecurity were cited as reasons for lack of association [33,67]. Notably, in Inuit populations, low-income individuals were more likely to consume the lower-cost, fortified food from government aid programs over the folate-poor traditional Inuit diet [111,112]. Hence, although these individuals were more likely to be food-insecure, their consumption of affordable fortified foods diminished the difference in folate deficiency between food-secure and food-insecure populations.

Vitamin D
None of the six studies that examined Vitamin D identified any association with food security [24,[56][57][58]83,89]. Several of these studies were conducted in South Asia and the Middle East, where dressing habits, especially for women, cover the skin and prevent adequate absorption of Vitamin D [58,100]. Since these practices occur independently of food security status, they may mask differences between food-secure and -insecure populations.

Vitamin B12
Three studies found no association between vitamin B12 and FI [31,41,67], and two studies found a negative association [51,79]. Only one study provided an explanation for the observed lack of association that was specific to vitamin B12 [41]. The authors suggested that due to religious beliefs prohibiting nonvegetarian foods, widespread B12 deficiency may have minimized the magnitude of difference in B12 levels between food-insecure and food-secure individuals. No specific explanations were provided by the studies which found negative associations between FI and B12.

General
In addition to biomarker-specific reasons, there were several broadly-applicable explanations repeated across studies for why limited associations were found between biomarker levels and food security status. First, subgroup analysis identified 40 instances of pediatric studies finding no relationship between food insecurity and biomarkers assessed (supplemental Figure S4). In many of the pediatric studies, household food security status may not have reflected the actual food security status of the child whose biospecimen was collected. Multiple study authors remarked that intrahousehold food allocation practices may shield children from the effects of food insecurity [40,56,57,67,73,88].
Second, it is possible that in a broadly impoverished population, biomarker differences between individuals with and without FI are reduced [58,81]. Difficulties with access to foods rich in specific nutrients may create a broad background of nutrient insufficiency that masks differences in nutrient biomarkers [51]. Likewise, an unexpectedly high prevalence of certain health conditions in a population, such as metabolic syndrome, may affect serum levels of certain biomarkers and once again mask differences by reducing statistical contrast [96]. Covariates such as socioeconomic status, education level, or occupation also play roles in mediating participant health behaviors, but were not routinely adjusted for and thus may have reduced signal strength within studies.
Finally, biomarker measurements were taken at a single time point, but food insecurity was diagnosed from eating patterns lasting weeks to a year [31]. Thus, if participants consumed meals different from their usual eating pattern before sample collection, their observed biomarker levels may not actually correlate to typical values when on their standard diet. People with food insecurity may also have cyclical eating patterns, where food consumption may be increased during some points and lower at others [87]. This has been reported in studies in people receiving SNAP. Multiple biomarker measurements may be necessary to account for these fluctuations.

Strengths and Limitations
A particular strength of this review is that it utilized data from physiologic biomarker levels. Studies measuring levels based on intake are subject to bias from participant recall, inexact estimates of nutrient content, and inter-individual differences in nutrient absorption. Nevertheless, this review also has multiple limitations. Due to its exploratory nature, data quality was not systematically assessed. Furthermore, due to the considerable heterogeneity in study characteristics, the review did not pool data or include a weighted analysis by population size. As such, only the ultimate directionality of a biomarker's association with food security status was tallied across studies. Inclusion criteria for acceptable food insecurity measurement tools were intentionally broad to allow for a comprehensive survey of the literature. Since the classification of participant food insecurity was carried out using a range of tools, the definition of food insecurity was not always consistent across studies.
Additionally, studies varied between measuring food security at the household versus individual level. Household food security was the more common measurement, but lacked the granularity to assess whether the participant whose biospecimen was being collected reflected the food security status of the household. This limitation may have been particularly relevant to studies in pediatric populations, since parents may protect their children from some of the effects of food insecurity, meaning that the household food insecurity status does not necessarily reflect the status of the child. Next, there are genetic components to certain deficiencies, and serum levels of nutrients are influenced by behaviors such as drinking and smoking and by conditions such as pregnancy. Serum levels of micronutrients may also be affected by supplementation, masking differences between populations. These variables were not always adjusted for between studies. Some of the included studies were conducted in highly specific populations with distinct dietary norms, which may limit the generalizability of conclusions to populations with different dietary practices. Finally, if only a subpopulation of a study sample showed a relationship between a biomarker and food security status, that study was still coded as demonstrating a relationship. This amplified the existence of potential relationships that may exist, but also reduced the data into a binary classification which may miss certain nuances.

Future Directions
As highlighted by the lack of consistent outcomes in existing studies, emphasis on single biomarkers to obtain an objective and quantitative assessment of dietary quality may be of limited value. Recent advances in nutritional untargeted metabolomics, when combined with traditional dietary assessments, may be an opportunity to develop shorter and more efficient nutrition status evaluation tools that combine abbreviated self-reported surveys with bioanalytic assessment [113][114][115]. As an adjunct to currently available dietary survey instruments, bioanalytic assessments can provide quantifiable estimates of dietary composition that are less prone to recall bias, reveal patterns of features that predict health outcomes, and identify metabolites associated with health outcomes that were not previously known (such as the recent example of identifying erythritol as a plasma biomarker associated with adverse cardiovascular events [116]). Such exploration may also capture nuances in health outcomes based on nutrient source-for example, it has been documented that for certain health outcomes, obtaining nutrients from supplements does not confer the same benefits as obtaining nutrients from food [117]. It should be noted though that many biomarkers are not specific to individual foods, so metabolomic quantification of exact dietary consumption is unlikely [14].
While bioanalytic methods offer increased objectivity, it is important to establish that they also face multiple limitations. First, pre-analytical handling has been shown to be a source of variation in metabolomic studies, with environmental exposure (e.g., light, oxidation), storage conditions (e.g., temperature, freeze-thaw cycles, preservatives), and timing of sample collection (e.g., circadian rhythm effects, effects of recent behaviors) all having been shown to affect reliability of result interpretation [118]. While generalizability of diet data captured from only one point in time is limited, it has nevertheless been demonstrated that multiple-sampling approaches, such as 24-h dietary recalls administered over three non-consecutive days, can approach the accuracy of more robust methods [119]. As such, it is plausible that repetitive biomarker measurements may provide more accurate estimates of habitual dietary patterns. Second, interindividual variation through genetic, environmental, and gut microbiota differences can confound findings. To overcome these effects, stratification of populations by "metabotype" may aid in signal detection [120]. Lastly, due to logistical constraints of large-scale sampling, metabolomic analyses may need to be conducted by various operators or laboratories, which generates batch effects that introduce artificial variation. While improved data-processing strategies help minimize these effects, they face compromises in statistical performance which may limit their utility [121].
Presently, food security measurement systems and policies focus more on food quantity than quality. Through exploration with bioanalytical methods, the dietary quality of FI individuals may be ascertained more accurately, and subcategories of FI that are reflective of nutrition security may be identified for more tailored intervention. Biomarker-based assessment methods will likely not supplant existing validated instruments for food security status assessment, but rather may serve as an adjunct to more holistically track outcome metrics of specific interventions and determine best strategies for combatting poor dietary quality in FI individuals.

Conclusions
Overall, the present review reveals limited scientific investigation into how biomarkers of dietary quality may differ between FI versus non-FI populations. Only three biomarkers of nutritional quality have been evaluated in this context. Of the 59 total biomarkers identified in this scoping review, few demonstrated consistent, directional relationships with food insecurity. Ultimately, future research is indicated to better characterize how nutrition security can be assessed in scalable, detailed formats for public health application.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/nu15163576/s1; Figure S1: Search queries for each of the utilized databases; Figure S2: Biomarkers associated with food insecurity among all non-pregnant adults; Figure S3: Biomarkers associated with food insecurity among US non-pregnant adults; Figure S4: Biomarkers associated with food insecurity among all children; Figure S5: Biomarkers associated with food insecurity among US children; Figure S6: Biomarkers associated with food insecurity among all US non-pregnant adults and children; Figure S7: Biomarkers associated with food insecurity among studies using NHANES data.