Systematic evidence and gap map of research linking food security and nutrition to mental health

The paper investigates how food security and nutrition (FSN) relate to common mental health outcomes across populations and settings. It frames the problem in the context of rising challenges: healthy diets remain unaffordable for many, contributing to undernutrition in LMICs and nutrition-related chronic diseases globally, while mental health disorders (particularly depression, anxiety, and stress) are leading contributors to disability worldwide. Despite a growing literature linking food insecurity, dietary patterns, nutrient status, anthropometry (notably BMI), and maternal mental health with nutritional outcomes in children and older adults, evidence is fragmented across topics, populations, measures, and study designs. The authors aim to systematically identify, classify, and map analytical studies linking diverse FSN constructs with mental health outcomes to create an interactive Evidence and Gap Map (EGM) that summarizes the scope of evidence, highlights concentrations and gaps, and guides future research and hypothesis generation.

Prior research shows: (a) food insecurity is often associated with poorer mental health, including depression, across contexts; (b) evidence linking specific nutrients to mental health is mixed or of variable quality; (c) dietary patterns and diet quality associate with depression and sometimes anxiety, though heterogeneity of measures limits inference; (d) relationships between BMI and mental health are long-studied, largely in HICs; (e) maternal mental health can be associated with poorer child diet diversity, micronutrient status, and growth in several, but not all, settings; and (f) in older adults, depression can contribute to nutritional risk and malnutrition. Existing systematic reviews are typically narrow—focused on specific populations or subsets of FSN and mental health indicators—and many primary studies are secondary analyses where FSN–mental health linkages are not the primary outcomes, limiting breadth and quality of evidence. This background underscores the need for a comprehensive mapping of the field to clarify where evidence is strong, where it is sparse, and to inform causal frameworks and future research priorities.

Design: Systematic identification and mapping of analytical studies linking FSN constructs to mental health outcomes into an Evidence and Gap Map (EGM). The team followed guidance for EGMs, employing expert-led screening and coding. An information specialist designed the search strategy using an index list of known literature. Data sources and search: Three databases were searched for studies published from 2000 through mid-2020 (search concluded halfway through 2020). Only English-language sources were searched; grey literature and non-English repositories were not included. Eligibility: Analytical studies (including observational, experimental, quasi-experimental, and systematic reviews/meta-analyses) that hypothesized relationships between FSN and mental health in either direction (FSN as exposure and mental health as outcome, or vice versa). Qualitative and mixed-methods studies were eligible. Exclusions included studies in populations with underlying health conditions (e.g., diabetes, CVD, HIV, TB, hospitalized patients) or niche groups (e.g., endurance athletes), and FSN measures not directly assessing food security, intake, or nutritional status (e.g., eating behaviors, stimulant foods, or breastfeeding intentions). Screening and selection: From 40,192 results, 30,896 records remained after deduplication and were title/abstract screened; 3,771 full texts were assessed; 1,945 studies met inclusion criteria and were coded and mapped. Coding captured FSN domains (food security/scarcity, nutritional risk/famine, diets, nutrient intakes, nutrient biomarkers, IYCF, birth outcomes, anthropometry), mental health domains (depression, anxiety, stress, mental well-being, hybrid domains), study design, hypothesis direction, geography (HIC/LMIC), and populations. Synthesis approach: The EGM provides descriptive mapping (counts, proportions, interactive filters) but does not pool effects or conduct formal quality assessment of all included studies. Figures include PRISMA flow, heat maps, Sankey diagrams, and geographic distributions.

Overall mapping: 1,945 studies were included. FSN measures by proportion: anthropometry 40%, diets 24%, nutrient intakes 14%, birth outcomes 13%, food scarcity 12%, nutrient biomarkers 10%, IYCF 6%. FSN details: - Anthropometry: BMI featured in 703 studies (36% of all mapped; ~90% of anthropometry studies). Some studies examined child stunting, wasting, and underweight (n=45 with depression; n=23 with hybrid domains; few with anxiety/stress, n=5). - Diets: Studies on dietary patterns/quality (16%) and specific food groups (12%) were prevalent. - Nutrient intakes (n=273): Macronutrients were most common (152/273; 94/152 on PUFA). Vitamins were next (110/273), focusing on B vitamins (65%), calcium (40%), and vitamin C (29%). Eighty-seven studies measured supplement intake. - Nutrient biomarkers (n=202): 55% vitamins, particularly vitamin D (66%), folate (25%), and B12 (20%). - Birth outcomes (n=245): Primarily birth weight (84%); many included multiple outcomes (31%). Nearly half linked birth weight with depression; an additional 35% with hybrid domains; only 28/245 studied mental health measures other than depression. - Food scarcity (n=230): 71% measured food security (commonly USDA scales; n=70 including adaptations). A few assessed famine exposure (n=9). Nutritional risk was mostly in older populations (n=70). - IYCF (n=124): Predominantly breastfeeding indicators (n=114). Mental health measurement: - Depression was assessed in 61% of studies; hybrid domains in 26%; stress 12%; mental well-being 12%; anxiety 10% (some studies included multiple measures). - Tools: 82% used screening questionnaires. Common tools included CES-D (n=332), EPDS (n=183), GDS (n=105), PHQ (n=104) for depression; GHQ (n=76), CBCL (n=41), HSCL (n=36), DASS (n=33) for hybrid domains; STAI for anxiety (n=64); PSS (n=84) and KSI (n=46) for stress; SF-36/RAND for well-being (n=83 of 229). - Diagnostic interviews were used in 9% of studies (mostly depression or hybrid); 8% used self-report diagnosis/medication/records. Only 14 qualitative or mixed-methods studies were identified. Reviews/meta-analyses: 89 on depression, 58 hybrid, 14 anxiety, 8 stress, 3 well-being. FSN–MH linkages (selected counts from Sankey/heat maps): - Anthropometry–depression: BMI with depression (n=401; 21% of all mapped) and with hybrid domains (n=192; 10%) and with well-being (n=109; 6%). - Diets–mental health: Diets linked with depression (n=278; 14%), hybrid (n=121; 6%), well-being (n=69; 4%). - Nutrient intakes: ~75% linked to depression; 18% to hybrid. PUFA intake linked to depression/hybrid in 89 studies; vitamin D intake and depression in 32 studies. Biomarkers showed similar patterns; vitamin D biomarker linked to depression in 66 studies. - Birth outcomes: ~50% focused on birth weight with depression; ~35% with hybrid; only 10 studies on fetal growth restriction with well-being/stress. - Food scarcity: Often linked to depression (e.g., nutritional risk in elderly: 56/70). Studies of food security more frequently included stress, anxiety, and well-being than other FSN groups (40% hybrid, 19% stress, 9% anxiety, 9% well-being alongside 72/163 on depression). - IYCF: Breastfeeding–depression examined in 91 studies; few on IYCF with anxiety (n=28), stress (n=11), or well-being (n=4). Child complementary feeding/diet diversity rarely linked to broader MH; only three studies compared any MH measure with child dietary diversity. Study methods and hypothesis direction: - Hypothesis direction: 66% (n=1291) treated FSN as exposure affecting mental health; 31% (n=600) treated mental health as exposure affecting FSN; 54 studies tested both directions over time (most on BMI and MH, n=31/54). - When MH was exposure (n=600): 39% examined MH→BMI (e.g., depression n=121; hybrid n=69; stress n=60; anxiety n=27; well-being n=9). Birth outcomes were the second-largest MH exposure group (147 total; 119 on birth weight). For IYCF outcomes (n=75), most were breastfeeding (n=67). Study designs: - Observational studies comprised 83% (46% cross-sectional; 37% longitudinal). Case-control 3%. - Reviews: 142 systematic reviews (48 with meta-analyses). Of 69 meta-analyses, 59 focused on depression or hybrid domains. - Experimental: 65 RCTs; 20 quasi-experimental (12 Mendelian randomization/genetic IV). Most RCTs tested nutrient intake exposures on MH outcomes (46/65), often supplements (38/47) including B vitamins, vitamin D (n=12 trials), zinc, and fatty acids (n=16 trials). Diet-pattern RCTs were fewer (e.g., Mediterranean, low fat/calorie, DASH, vegetarian); only three trials were MH interventions with FSN outcomes. Geography and populations: - By region/political category (number of studies): HIC 1489; LMIC 446; North America 549; Europe 521; Asia 418; Oceania 160; Global 160; Africa 81; South America 67; Arab Countries 24; Central America 18. - 23% were set in LMICs; evidence concentrated in the US, UK, Australia, Europe, and Asia; African, Latin American, and Arab country evidence was comparatively sparse. - HIC vs LMIC: proportionally more nutrient intake studies in HIC (15% vs 9%); more food scarcity studies in LMIC (18% vs 10%). BMI studies were more prevalent in HIC (95% of 611 anthropometry studies) vs LMIC (72% of 172); LMIC had more child relative height (20% vs 1%) and relative weight (11% vs 2%) measures. 82% of studies using validated diagnostic MH tools were from HIC. Anxiety, stress, and well-being measures were more common in HIC (e.g., well-being 13% vs 7%). - Populations: 49% general/representative adults; 21% mid-to-later-life. Among child-including studies (n=695): under-5s in 433, ages 5–12 in 221, adolescents 13–18 in 248. Under-5 MH measurement was uncommon (n=106) compared to FSN measurement (n=423). Pregnant/perinatal women, mothers, and fathers featured in 28% of studies; studies of women-only (non-perinatal) were 8% (n=158); men-only n=42 (2%). Cross-cutting analyses frequently examined parental MH and child FSN (n=355), mostly mothers; fathers were rarely included (n=8). Household food security and individual MH were linked in 107 studies (often adults and perinatal women). Time trends: The number of analytical studies linking FSN and MH increased steadily from 2000 to 2020 (drop in 2020 due to mid-year search cutoff).

The mapping demonstrates a robust but uneven evidence base. Research is heavily concentrated on depression and BMI, with substantial work on diets, nutrient intakes (especially supplements), and vitamin D biomarkers. In contrast, relationships involving anxiety, stress, mental well-being, IYCF practices, and child growth are comparatively under-studied, despite plausible pathways (e.g., food insecurity may elevate worry and stress). Experimental and quasi-experimental evidence is limited and mainly focused on nutrient supplementation rather than broader FSN constructs or mental health interventions with FSN outcomes, constraining causal inference. Reviews and meta-analyses exist but are fewer relative to the volume of observational studies, and many reviews lack meta-analyses. Geographically, evidence is dominated by HICs, with LMIC studies concentrated in a few countries and relatively sparse representation of poorer or more rural contexts where food insecurity and undernutrition burdens are highest. Populations most often studied include general adults and mid-to-later-life groups; perinatal women and mothers are frequently examined as MH exposures affecting child FSN outcomes, whereas paternal roles and reciprocal pathways (child FSN influencing parental MH) are rarely investigated. Cross-population linkages largely focus on maternal mental health and child nutrition/growth. Overall, while associations between FSN and mental health are well-documented in several domains (e.g., BMI–depression, diet quality–depression, vitamin D–depression), there are notable gaps in domains, populations, geographies, and methods needed to elucidate mechanisms and inform policy and programming.

This study provides the first comprehensive Evidence and Gap Map of analytical studies linking food security and nutrition to mental health across measures, populations, study designs, and settings. It synthesizes 1,945 studies, showing concentrations of evidence around depression, BMI, diet quality/patterns, and specific nutrients (notably vitamin D and PUFA), and identifies substantial gaps in anxiety, stress, well-being, IYCF, child growth, LMIC contexts, paternal involvement, and causal/experimental designs. The interactive EGM serves as a resource to navigate the literature, identify priority gaps, and guide the design of primary research and evidence syntheses. Future research should prioritize: diversified mental health outcomes beyond depression; understudied FSN domains (IYCF, child growth, micronutrients like selenium and antioxidants); rigorous experimental and quasi-experimental studies (including diet-pattern and food security interventions); qualitative and mixed-methods to elucidate mechanisms; and broader geographic and population representation, especially in diverse LMIC settings and inclusive of fathers and family dynamics.

Key limitations include: (1) scope and synthesis—EGMs map but do not estimate pooled effects; comprehensive quality assessment and meta-analysis of nearly 2,000 studies was beyond scope; (2) search constraints—limited to publications from 2000 onward, English-language sources, and selected databases; grey literature and non-English repositories were excluded; (3) eligibility filters—excluded studies in populations with underlying health conditions or niche groups to reduce confounding, which may omit relevant evidence; (4) measurement issues—heterogeneity of FSN and mental health measures, and challenges disentangling mental well-being from broader quality-of-life constructs; (5) evidence imbalances—dominance of cross-sectional/observational designs and HIC settings limits causal inference and generalizability to diverse LMIC contexts.