Food insecurity (FI) is characterised by the US Department of Agriculture as limited or uncertain access to nutritionally adequate and safe foods or the ability to acquire acceptable foods in socially acceptable ways due to limited financial resources(Reference Anderson1). FI represents a staggering global burden(Reference Lee2). The prevalence of moderate or severe FI has substantially increased from 8·3 % in 2014 to 25·9 % in 2020, with a projected two billion individuals at risk of hunger(3). The condition is associated with an increased risk of morbidity, including type 2 diabetes(Reference Tait, L’Abbe and Smith4), CVD(Reference Liu and Eicher-Miller5), anaemia(Reference Moradi, Arghavani and Issah6), metabolic syndrome(Reference Park and Strauss7,Reference Moradi, Mirzababaei and Dadfarma8) , stunting(Reference Moradi, Mirzababaei and Dadfarma8), mental disorders(Reference Pourmotabbed, Moradi and Babaei9) and mortality(Reference Sun, Liu and Rong10), highlighting the need for preventative and management strategies.
Sleep plays a significant role in the physical and mental health status(Reference Pourmotabbed, Boozari and Babaei11,Reference Pourmotabbed, Ghaedi and Babaei12) . Many factors are known to affect the quality and quantity of sleep, including ageing(Reference Madrid-Valero, Martinez-Silva and Couto13), chronic diseases(Reference Visvalingam, Sathish and Soljak14), obesity(Reference Sa, Choe and Cho15), occupational stress(Reference Deng, Liu and Fang16), poor sleep environment(Reference Xiong, Lan and Lian17), smoking(Reference Kieliszek and Lipinski18), excessive caffeine(Reference Snel and Lorist19), alcohol(Reference Inkelis, Hasler and Baker20) and drinking consumption. Similarly, FI may contribute to impaired sleep behaviours(Reference St-Onge, Mikic and Pietrolungo21), albeit the evidence is less conclusive.
In general, available studies have reported positive associations between FI and poor sleep quality; however, results are less consistent on select sleep health outcomes, including sleep quality and duration(Reference Ding, Keiley and Garza22–Reference El Zein, Shelnutt and Colby27). To that end, Ding et al.(Reference Ding, Keiley and Garza22) and Jordan et al.(Reference Jordan, Perez-Escamilla and Desai25) have reported associations between FI and poor sleep quality across the mild to severe status. However, Grandner et al.(Reference Grandner, Chakravorty and Perlis28) showed only extreme levels of FI are related to poor sleep quality, unlike mild levels. Consistently, data on associations between FI and sleep duration are mixed. Troxel et al.(Reference Troxel, Haas and Ghosh-Dastidar29) reported that higher levels of FI were associated with a greater risk of short or long sleep duration compared with normal sleep duration (7–9 h). Similarly, Narcisse et al.(Reference Narcisse, Long and Felix30) and Jordan et al.(Reference Jordan, Perez-Escamilla and Desai25) showed that FI is associated with short sleep duration. In contrast, they reported a lack of association between FI and long sleep duration(Reference Jordan, Perez-Escamilla and Desai25,Reference Narcisse, Long and Felix30) . Conversely, Whinnery et al.(Reference Whinnery, Jackson and Rattanaumpawan31) exhibited FI is associated with both short and long sleep duration. Collectively, little can be concluded on the direction and magnitude of the relationship between FI and sleep behaviours.
To our knowledge, no meta-analysis has pooled evidence on the relationship between FI and sleep behaviours, despite a clear need. To address this knowledge gap, we conducted a systematic review and meta-analysis of observational studies to delineate associations between FI and the quality and quantity of sleep in adults (≥18 years). We also comprehensively evaluated factors that may influence these associations, including the severity of FI and biological and socio-demographic characteristics of study populations (e.g. age, sex, BMI, race, ethnicity, mental health, education and income status).
Methods
Systematic search and study selection
The work presented herein was carried out according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA)(Reference Page, McKenzie and Bossuyt32). The study protocol was registered at the PROSPERO (Prospective Register of Systematic Reviews; registration identifier: CRD42021275645). A comprehensive literature search was conducted using the databases of PubMed, Scopus, Web of Science and Embase from inception until 6 June 2022. The search was performed using following medical subject heading (MeSH) and defined search terms without any language or date restrictions: ((‘Food Supply’[Mesh] OR ‘Food Supply’[Title/Abstract] OR ‘Food Supplies’[Title/Abstract] OR ‘FI’[Title/Abstract] OR ‘Food Insecurities’[Title/Abstract] OR ‘Food security’[Title/Abstract] OR ‘Food securities’[Title/Abstract]) AND (‘Sleep’[Mesh] OR ‘Sleep’[tiab] OR ‘insomnia’[tiab] OR ‘insomnias’[tiab] OR ‘sleep problems’[tiab] OR ‘sleep quality’[tiab] OR ‘sleep duration’[tiab] OR ‘sleep deprivation’[tiab] OR ‘sleep disturbance’[tiab] OR ‘sleep disorders’[tiab]) (see online Supplemental Table 1). Also, the references of retrieved records were evaluated manually to identify relevant citations for inclusion in our literature search.
Eligibility criteria
Studies were included in the final analysis if they: (1) were observational (cross-sectional, cohort and case–control); (2) were conducted on adults (≥18 years); and (3) reported effect estimates in the form of OR, relative risk, or hazard ratio and with corresponding 95 % CI on associations between FI and sleep quality across short (≤6 h) or long (≥9 h) durations.
Exclusion criteria were studies: (1) that conducted on children or adolescents (<18 years); (2) that had insufficient data for inclusion in our analyses; and (3) that had with inappropriate designs, including interventional studies, reviews, letters, editorials, conference proceedings, notes or surveys.
Study selection
The titles and abstracts of all identified records in our literature search were assessed independently by two authors (SM and M-AH-K), followed by a full-text review of eligible records. All discrepancies were resolved by consensus with a third investigator (HM).
Data collection
Extracted data for each included study were as follows: (1) the first author’s name; (2) study publication year; (3) databases used; (4) study design; (5) study country of origin; (6) study sample size; (7) biological and socio-demographic characteristics of study populations (e.g. age, sex, BMI, race, ethnicity, mental health, education and income status where available); (8) level of FI severity; (9) assessment methods for FI and sleep behaviours; (10) relevant effect estimates; (11) study main findings; and (12) any adjusted analyses. Two investigators (SM and HM) independently extracted data for all included records using a standard information extraction template. Data extraction was reviewed by all other authors (S-NM, ST, MK, S-MG, S-PM, PA, SN-M and WM) for any potential extraction error.
Quality assessment
Two investigators (SM and HM) independently examined the quality of each included study using the Newcastle–Ottawa scale(Reference Modesti, Reboldi and Cappuccio33). The method of quality evaluation has been formerly described(Reference Moradi, Arghavani and Issah6,Reference Pourmotabbed, Ghaedi and Babaei12) .
Statistical analysis
All statistical tests were conducted using STATA (version 14.0; Stata Corp.). To analyse associations between FI and sleep behaviours, fully adjusted risk estimates for poor sleep quality and short or long sleep duration were pooled. Pooled OR and 95 % CI were estimated using a weighted random-effects model per the DerSimonian–Laird approach(Reference DerSimonian and Laird34). The heterogeneity among the studies was examined by the Cochran Q and I 2 statistics (I 2 = (Q-df)/Q × 100 %; I 2 < 25 %, no heterogeneity; I 2 = 25–50 %, moderate heterogeneity; I 2 = 50–75 %, considerable heterogeneity, I 2 > 75 %, extreme heterogeneity). The heterogeneity was considered significant if the Q statistic had P < 0·1 or I 2 > 50 %. To identify the sources of heterogeneity, subgroup analyses were conducted based on FI levels (mild, moderate and severe)(Reference Coates, Swindale and Bilinsky35), sleep problems (trouble falling asleep and difficulty staying asleep) and country (USA and Mexico). We also performed subgroup analyses based on age (<50 and ≥50 years), ethnicity/race (mixed and Latino) and number of participants (<4000 and >4000). Our subgroup analyses were justified based on eight recommended criteria of the Instrument to Evaluate the Credibility of Effect Modification Analyses (ICEMAN)(Reference Baker, White and Cappelleri36). We also performed meta-regression analyses to evaluate the link between the risk of sleep quality or quantity and heterogeneity between studies. Further, we performed sensitivity analyses by removing each study and recalculating the overall effect size to determine whether an individual study exerted undue influence. Funnel plots and results of Begg’s and Egger’s tests were used to assess publication bias. Results were considered significant at P < 0·05.
Results
The systematic search resulted in 651 records (Fig. 1), of which 318 records were screened after removing duplicates. Of these 318 records, 302 were excluded because they did not meet our inclusion criteria, resulting in nineteen eligible studies for full-text evaluation(Reference Ding, Keiley and Garza22–Reference Whinnery, Jackson and Rattanaumpawan31,Reference Bermúdez-Millán, Perez-Escamilla and Segura-Perez37–Reference Widome, Jensen and Bangerter42) . Of these nineteen studies, ten were excluded because they were conducted on children(Reference Na, Eagleton and Jomaa40) or adolescents(Reference Wang43), provided insufficient data for inclusion in our analyses(Reference Bermúdez-Millán, Perez-Escamilla and Segura-Perez37–Reference Liu, Njai and Greenlund39,Reference Richards and Specker41,Reference Widome, Jensen and Bangerter42,Reference Gyasi, Asamoah and Gyasi-Boadu44,Reference Cheng, Luo and Perkins45) , or used variable domains to measure FI(Reference Isaura, Chen and Su24). Together, nine eligible studies (n 83 764)(Reference Ding, Keiley and Garza22,Reference Hagedorn, Olfert and MacNell23,Reference Jordan, Perez-Escamilla and Desai25–Reference Whinnery, Jackson and Rattanaumpawan31) were included in our study (Fig. 1).
Fig. 1 PRISMA flow diagram of the study
All included studies had a cross-sectional design (Table 1) and were published between 2013 and 2021 and conducted in the USA(Reference Ding, Keiley and Garza22,Reference Hagedorn, Olfert and MacNell23,Reference Nagata, Palar and Gooding26–Reference Whinnery, Jackson and Rattanaumpawan31) and Mexico(Reference Jordan, Perez-Escamilla and Desai25). OR on the link between FI and quality and quantity of sleep were pooled across these nine studies for meta-analyses. Seven studies assessed poor sleep quality risk (n 47 439), and four reported sleep duration risk (n = 29 583).
Table 1 Summary of studies included in the meta-analysis
KNHANES, Korea National health and nutritional examination survey; ENSANUT-2012, The 2012 Mexican national health and nutrition survey; ELCSA, The Latin American and Caribbean food security scale; PSQI, Pittsburgh sleep quality index; IFLS5, Indonesian family life survey; PROMIS, patient-reported outcomes measurement information system; CES-D, epidemiologic studies – depression.
Results of the study quality assessment for each study are shown in Table 1. Briefly, quality assessments revealed that seven studies had high quality(Reference Ding, Keiley and Garza22,Reference Hagedorn, Olfert and MacNell23,Reference Jordan, Perez-Escamilla and Desai25–Reference El Zein, Shelnutt and Colby27,Reference Troxel, Haas and Ghosh-Dastidar29,Reference Narcisse, Long and Felix30) and two had medium quality(Reference Grandner, Chakravorty and Perlis28,Reference Whinnery, Jackson and Rattanaumpawan31) (Table 1).
Sleep quality
FI was associated with an increased risk of poor sleep quality in adults (OR = 1·45; 95 % CI (1·24, 1·70), P < 0·001, n 7; Fig. 2). Studies were highly heterogenous (I 2 = 95 %, P < 0·001).
Fig. 2 Forest plot showing the OR and 95 % CI of the association between food insecurity and the risk of poor sleep quality
Subgroup analysis showed increased risk of poor sleep quality corresponding to the severity of FI across mild (OR = 1·31; 95 % CI (1·16, 1·48), I 2 = 0 %, P < 0·001, n 5), moderate (OR = 1·49; 95 % CI (1·32, 1·68), I 2 = 0 %, P < 0·001, n 5) and severe (OR = 1·89; 95 % CI (1·63, 2·20), I 2 = 0 %, P < 0·001, n 5) levels (Table 2). Similarly, subgroup analysis by sleep problems showed that FI was associated with an increased the risk of trouble falling asleep (OR = 1·39; 95 % CI (1·05, 1·83), I 2 = 91 %, P = 0·002, n 3) and trouble staying asleep (OR = 1·91; 95 % CI (1·37, 2·67), I 2 = 89 %, P < 0·001, n 3; Table 2). Also, subgroup analysis based on country revealed that FI was associated with an increased the risk of poor sleep quality across studies conducted within (OR = 1·44; 95 % CI (1·22, 1·70), I 2 = 95 %, P < 0·001, n 6) or out of (OR = 1·55; 95 % CI (1·36, 1·77), n 1) USA.
Table 2 Subgroup analysis to assess the associations between food insecurity and the quality and quantity of sleep
* Calculated by random-effects model.
† P-value for heterogeneity within the subgroup.
‡ P-value for heterogeneity between subgroups using meta-regression analyses.
To further explore the sources of heterogeneity, meta-regression analyses were conducted to identify any influence of FI degree, sleep problems, age, race/ethnicity, study sample size and adjusted risk estimates across different exposure categories (Table 3). Heterogeneity was decreased following meta-regression analyses based on FI levels (P = 0·003, I 2 = 0 %; Table 3 and see online Supplemental Fig. 1). However, sleep problems, age, race/ethnicity, number of participants, and studies that controlled for sex, BMI, mental, education, and income did not explain the sources of heterogeneity.
Table 3 Findings from meta-regressions
Sleep quantity
FI was associated with an increased risk of short (OR = 1·14; 95 % CI (1·07, 1·21), P < 0·001, n 4) or long (OR = 1·14; 95 % CI (1·03, 1·26), P = 0·010, n 4) and sleep duration (Figs. 3 and 4, respectively), and studies were homogenous (All: I 2 = 0 %; All: P ≥ 0·05). In addition, subgroup analysis showed that a severe level of FI is associated with an increased risk of short sleep duration (OR = 1·59; 95 % CI (1·16, 2·18), I 2 = 46 %, P = 0·004, n 2; Table 2). In contrast, meta-regression analyses based on pooled FI levels, age, race and study size could not explain the sources of heterogeneity (All: P > 0·05, Table 3 and see online Supplemental Figs. 2 and 3).
Fig. 3 Forest plot showing the OR with 95 % CI of the association between food insecurity and the risk of short sleep duration
Fig. 4 Forest plot showing OR with 95 % CI of the association between food insecurity and the risk of long sleep duration
Sensitivity analysis and publication bias
Sensitivity analysis revealed that the pooled effect estimates were not affected by any single study included in our analyses. The Egger’s test (P = 0·01) revealed a publication bias for studies assessing the relationship between FI and the risk of poor sleep quality. However, the bias was not evident using Begg’s test results (P = 0·71) or a symmetric funnel plot (Fig. 5(a)). Further, we observed no publication bias in studies evaluating the link between FI and short (P = 0·30 for Begg’s test and P = 0·39 for Egger’s test; Fig. 5(b)) and long (P = 0·49 for Begg’s test and P = 0·73 for Egger’s test; Fig. 5(c)) sleep duration.
Fig. 5 Funnel plot for evaluation publication bias in studies reporting OR and 95 % CI of the association between food insecurity and risk of poor sleep quality (a), short sleep duration (b), and long sleep duration (c)
Discussion
Few studies have examined the relationship between FI and non-nutritional health outcomes, including sleep behaviours. To our knowledge, the present work is the first to investigate associations between FI and the quality and quantity of sleep. The most significant finding of our study was that FI was associated with an increased risk of poor sleep quality in adults. Also, FI was associated with an increased risk of short and long sleep duration. Together, our findings highlight the adverse influence of FI on sleep behaviours.
Our observations add a novel dimension to current evidence about the negative influence of FI on sleep health status in the general adult population and extend previous reports. Our results are consistent with those of a cross-sectional study on patients with type 2 diabetes, and Bermúdez-Millán et al. (Reference Bermúdez-Millán, Perez-Escamilla and Segura-Perez46) demonstrated that household FI is a common and potent household stressor related to suboptimal sleep quality through psychological distress. Similarly, Liu et al. corroborated associations between FI, frequent mental distress and insufficient sleep among adults across twelve states in the USA.(Reference Liu, Njai and Greenlund39) Consistently, Pinto et al. (Reference Pinto and Bertoluci47) reported that FI is associated with increased odds (OR: 2·25; 95 % CI (1·11, 4·55)) of poor sleep quality in children.
The biological and psychosocial factors involved in mechanisms behind the association between FI status and adult sleep behaviours are less clear. However, this relationship may be mediated, at least partially, through mental health disorders (e.g. depression or depressive symptoms)(Reference Silverman, Kriegar and Kiefer48). FI is associated with an increased risk of depression(Reference Heflin, Siefert and Williams49–Reference German, Kahana and Rosenfeld51), anxiety(Reference Hadley and Patil52) or stress(Reference Hamelin, Habicht and Beaudry53,Reference Seligman, Laraia and Kushel54) . These mental health complications are known to be associated with adverse sleep quality(Reference Zou, Wang and Sun55,Reference João, Jesus and Carmo56) . Individuals affected by FI present with perceived powerlessness, disappointment, embarrassment and guilt, which may contribute to anxiety and depressive symptoms(Reference Whittle, Palar and Seligman57). Furthermore, those with FI are more likely to consume convenience foods that are usually high in fat and refined sugars and are subsequently linked with poorer mental health through mechanisms explained in greater detail previously(Reference Moradi, Mirzababaei and Dadfarma8,Reference Lang, Beglinger and Schweinfurth58,Reference Lane, Davis and Beattie59) . Stress and depression may also exacerbate FI status secondary hormonal imbalance, including aggravated cortisol secretion and dysregulation of the hypothalamic–pituitary–adrenal axis(Reference Pourmotabbed, Moradi and Babaei9). These alterations have been known to disrupt sleep(Reference Coplan, Gupta and Karim60,Reference Vgontzas and Chrousos61) . Sleep disturbances can, in turn, alter appetite regulation medicated by increasing ghrelin and decreasing leptin levels. The compensatory mechanism of leptin reduces appetite and increases energy expenditure through the hypothalamic receptors(Reference Frank, Gonzalez and Lee-Ang62). Also, low leptin levels have been associated with poor sleep quality and a propensity for depressive symptoms(Reference Frank, Gonzalez and Lee-Ang62). Presently, the relative contributions of these individual factors to sleep behaviours are less conclusive by robust evidence, pointing to a research gap
This study has some strengths, including a comprehensive search strategy. This is the first meta-analysis to report associations between FI and sleep quality and quantity. Most studies included in our meta-analysis accounted for critical confounding factors. We performed several subgroup analyses to determine the source of the heterogeneity. However, important limitations should be acknowledged in the interpretation of our findings. Our work included cross-sectional studies. Therefore, no causality may be inferred on the link between FI and sleep behaviours. Most included studies relied on self-reported measures for FI and sleep. Accordingly, our observations are likely prone to over- or under-estimations of these measures secondary to the recall bias. Most (eight) studies were conducted in the USA; therefore, our findings may not be generalisable to low- and middle-income countries. Moreover, we observed considerable variability across studies in methods (e.g. surveys) used to measure FI and sleep outcomes, possibly contributing to measurement errors and a misclassification bias, which have been corroborated in systematic reviews and meta-analyses of this type(Reference Kazemi, Hadi and Pierson63,Reference Kazemi, Kim and Wan64) . Also, the assessment of FI and sleep behaviours occurred in different years, making it challenging to detect whether FI levels remained unchanged when sleep behaviours were evaluated. We observed significant heterogeneity among included studies for sleep quality. The heterogeneity was attenuated when the meta-analysis was subgrouped by the level of FI, and these results were approved following meta-regression analyses. However, other factors include sleep problems, country, age, race, number of participants and study adjustments did not explain the sources of heterogeneity. Our observations highlight the need for further research to elucidate the underlying factors and mechanisms that could explain the link between FI and poor sleep behaviours.
Conclusions
The present meta-analysis of observational studies revealed that FI was associated with poor sleep quality and quantity in adults. Our observations extend the growing evidence on associations between FI and physical and mental health. Findings from the present work highlight the need for preventative and management strategies that address FI and sleep behaviours. Future well-designed longitudinal studies with larger sample sizes should confirm our observations.
Acknowledgements
Acknowledgements: None. Financial support: This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors. Authorship: S.M. designed this study. S.M. and H.M. conducted the literature search. S.M. and M.-A.H. performed the statistical analysis and interpretation of the data. S.M., S.G.H. and S.N. wrote the manuscript. W.M., S.-N.M., S.T. and M.K. critically revised the manuscript. All authors approved the final version of the manuscript. Ethics of human subject participation: Not applicable.
Conflicts of interest:
The authors declare that they have no conflict of interest.
Supplementary material
For supplementary material accompanying this paper visit https://doi.org/10.1017/S1368980022002488
Open access
