Associations between food insecurity in high-income countries and pregnancy outcomes: A systematic review and meta-analysis

Background Maternal nutrition is crucial for health in pregnancy and across the generations. Experiencing food insecurity during pregnancy is a driver of inequalities in maternal diet with potential maternal and infant health consequences. This systematic review explored associations between food insecurity in pregnancy and maternal and infant health outcomes. Methods and findings Searches included 8 databases (MEDLINE, Embase, Scopus, Web of Science, PsychInfo, ASSIA, SSPC in ProQuest, and CINAHL), grey literature, forwards and backwards citation chaining, and contacting authors. Studies in high-income countries (HICs) reporting data on food insecurity in pregnancy and maternal or infant health, from January 1, 2008 to November 21, 2023 were included. Screening, data extraction, and quality assessment were carried out independently in duplicate. Random effects meta-analysis was performed when data were suitable for pooling, otherwise narrative synthesis was conducted. The protocol was registered on PROSPERO (CRD42022311669), reported with PRISMA checklist (S1 File). Searches identified 24,223 results and 25 studies (n = 93,871 women) were included: 23 from North America and 2 from Europe. Meta-analysis showed that food insecurity was associated with high stress level (OR 4.07, 95% CI [1.22, 13.55], I2 96.40%), mood disorder (OR 2.53, 95% CI [1.46, 4.39], I2 55.62%), gestational diabetes (OR 1.64, 95% CI [1.37, 1.95], I2 0.00%), but not cesarean delivery (OR 1.42, 95% CI [0.78, 2.60], I2 56.35%), birth weight (MD −58.26 g, 95% CI [−128.02, 11.50], I2 38.41%), small-for-gestational-age (OR 1.20, 95%, CI [0.88, 1.63], I2 44.66%), large-for-gestational-age (OR 0.88, 95% CI [0.70, 1.12] I2 11.93%), preterm delivery (OR 1.18, 95% CI [0.98, 1.42], I2 0.00%), or neonatal intensive care (OR 2.01, 95% CI [0.85, 4.78], I2 70.48%). Narrative synthesis showed food insecurity was significantly associated with dental problems, depression, anxiety, and maternal serum concentration of perfluoro-octane sulfonate. There were no significant associations with other organohalogen chemicals, assisted delivery, postpartum haemorrhage, hospital admissions, length of stay, congenital anomalies, or neonatal morbidity. Mixed associations were reported for preeclampsia, hypertension, and community/resilience measures. Conclusions Maternal food insecurity is associated with some adverse pregnancy outcomes, particularly mental health and gestational diabetes. Most included studies were conducted in North America, primarily the United States of America, highlighting a research gap across other contexts. Further research in other HICs is needed to understand these associations within varied contexts, such as those without embedded interventions in place, to help inform policy and care requirements.


Introduction
Food insecurity is "the limited or uncertain availability of nutritionally adequate or safe foods and limited or uncertain ability to acquire foods in socially acceptable ways"AU : PerPLOSstyle; italicss [1].Poverty, unemployment, and low income are key drivers of food insecurity in both low-and middleincome countries (LMICs) and high-income countries (HICs) [2].Further, the double burden of malnutrition with increasing rates of overweight, obesity, and noncommunicable diseases is also shared.However, it is important to consider HICs independently due to the vast differences not only in the food system, food supply, and food environment, but for many HICs, food insecurity was relatively hidden until after the 2008 global financial crisis and parallel rise in charitable food aid [3].The global recession that followed led to increased poverty rates and unemployment in HICs [4,5].Social and health inequalities have been further exacerbated from the Coronavirus pandemic, war, and ongoing cost-of-living crises meaning increased costs of food and energy prices [6,7].Concurrently, lack of routine population measurement and use of noncomparable measurements has prevented accurate understanding of prevalence of food insecurity [8].A study reporting prevalence in HICs between 2014 and 2018 estimated the prevalence of moderate or severe food insecurity as 6.5%, although there was a wide distribution at country level ranging from around 3% to 15% [9].The study also reported sociodemographic patterns in prevalence, with age, gender, household occupancy, and income being important.More recently, the United Nations estimated that food insecurity doubled from the start to the end of 2020, impacting 135 to 265 million people worldwide [10,11].During that period, the Coronavirus pandemic highlighted clear inequalities in access to food and health outcomes both between and within countries [12].Ongoing since 2021, HICs have been experiencing a cost-of-living crisis, further pushing households experiencing financial hardship into poverty [7].
In HICs, food insecurity is a major public health concern given its wide-reaching significant impact on diet, nutrition, physical and mental health [13].It is therefore important to understand the impact of experiencing food insecurity during critical periods of nutrition and development crucial for health across the generations [14].One critical period is during the first 1,001 days of life (from conception until a child is 2 years of age), particularly during pregnancy.Suboptimal nutrition during pregnancy can contribute to poorer maternal and infant health outcomes [14].During pregnancy and breastfeeding, women have more complex nutritional requirements, with increased potential health impacts from food insecurity [15,16].Further, women are particularly vulnerable to experiencing food insecurity due to increased likelihood of working lower-paid, part-time jobs while maintaining caring responsibility of children and elderly within the family unit and/or heading a single parent household [17][18][19].Within households, experiences of food insecurity may be unevenly distributed with women more affected than men [20].Qualitative research shows women tend to be household food managers, thus experiencing the psychosocial costs from negotiating access to sufficient healthy food, often foregoing food to ensure others in their household are fed [21][22][23].This increase in women's susceptibility to food insecurity could have consequences for pregnancy physical and mental health outcomes.Qualitative research reports maternal changes to food behaviours such as restricting eating patterns, reducing portion size, skipping meals, and forgoing food.Further, a recent review exploring associations between maternal food insecurity and pregnancy weight and diet reported poorer diet quality such as reduced consumption of fruit and vegetables, increased red processed meats, and lower intakes of vitamin E, and significantly increased odds of maternal obesity [24].Maternal obesity is highly prevalent in HICs and increases pregnancy complications such as large-for-gestational age (LGA) babies, preeclampsia, gestational diabetes mellitus (GDM), and in the longer-term type 2 diabetes [25,26], increased depression [27,28].The recent review did not report associations between food insecurity and other pregnancy outcomes.Observational studies have found maternal food insecurity to be associated with poor maternal mental health and be a risk factor for adverse pregnancy and birth outcomes [29][30][31]; however, there is an absence of systematic reviews reporting meta-analysis of these associations in HICs contexts.
This systematic review aimed to explore the relationship between food insecurity and pregnancy outcomes relating to maternal physical and mental health, and infant health.The review is set within the context of the global recession from 2008, when HICs experienced increasing prevalence of poverty and food insecurity.

Methods
The systematic review was registered on PROSPERO (CRD42022311669) and is reported as per the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [32] (S1 File) supplemented with MOOSE (Meta-analyses Of Observational Studies in Epidemiology) Checklist guidelines specific to observational studies [33] (see S1 File).
Database searches were completed November 21, 2023.A comprehensive search strategy was developed using search terms relating to "food insecurity" (exposure), "pregnancy" (population), and "observational" (study design) with Boolean operators (see S1 Table ).We did not include search terms relating to outcomes, as the aim was to be holistic and include studies reporting any pregnancy outcomes specific to maternal and infant health.No search terms were included relating to HICs, and studies in LMIC contexts were excluded during the screening stages.The search strategy was applied to MEDLINE, Embase, PsycInfo, Scopus, Web of Science, Applied Social Science Index, and Abstracts (ASSIA), Social Sciences Premium Collection (SSPC) in ProQuest and Cumulative Index to Nursing and Allied Health Literature (CINAHL) from January 1, 2008 to November 21, 2023.To limit the potential impact of publication bias, additional searches were conducted with searches for grey literature carried out using Trove, Open Access Theses and Dissertations (OATD) and stakeholder websites (see S2 Table ).We also conducted forward and backwards citation chaining for all included studies using citationchaser (an R package and Shiny app [34]).We contacted authors when additional data was required to inform the screening or to include in meta-analysis (see S3 Table ).Supplementary searches and contacting authors were completed in March 2024.
A modified version of the Population, Intervention, Control, Outcome, Study design (PICOS) framework: Population, Exposure, Comparison, Outcome, Study design (PECOS) [35] was used to develop eligibility criteria.The population (P) was pregnant women living in HICs (GNI per capita �$12,696 [36]).The exposure (E) was an explicit measure of food insecurity, and the comparison group (C) was food security/no food insecurity.Eligible study designs (S) were quantitative observational studies and primary observational data sources from grey literature.A date restriction was applied and aligned with the global financial crises which occurred in 2008.We excluded any studies where the data were primarily collected prior to 2008.There were no restrictions on reporting language to limit bias [37].
The search results were deduplicated in EndNote 20 [38] and Rayyan [39].Title and abstract screening were conducted in Rayyan by a team of 7 reviewers independently and in duplicate.Full-text screening for any potentially eligible studies was also carried out independently and in duplicate.Conflicts between reviewers screening decisions were resolved by group discussion.We contacted corresponding authors when required, including following up on published conference abstracts for any subsequent publication of full papers, or when there was missing information in the publication essential for the screening decisions (e.g., data collection time period).Data extraction and quality assessment were carried out independently in duplicate, with conflicts resolved by group discussion.The data extraction template included: type of publication, study type, country of study, data collection time period, study aim, measurement and definition of food insecurity, study participant information, and study outcomes.The Newcastle-Ottawa scale (NOS) cohort study tool was used to assess the quality of the included studies [40].This scale assesses information and selection bias, for example, relating to participant identification, representativeness, and measurement of exposures and outcomes, and confounding variables which are the key factors to consider when assessing quality and bias in observational study designs.A quality scoring system was applied based on the number of "stars" awarded: 0 to 2 was categorised as low quality, 3 to 5 stars medium quality, and 6 to 8 stars high quality (see S4 Table ).
Meta-analysis was carried out when there were at least 3 studies reporting data suitable for pooling.If odds ratios (ORs) were not reported in the included papers, then they were calculated using reported frequency data or by contacting authors for the required data, with the aim of increasing the number of studies to pool in a meta-analysis.Summary ORs were calculated using the random effects model using Stata/MP version 18 [41,42].Heterogeneity between studies was explored using I 2 [43], and >75% was considered significant heterogeneity [44].The protocol included a plan to conduct meta-regression, sensitivity analysis, Eggers tests, and funnel plots for all meta-analyses; however, as no meta-analysis included 10 or more studies, these were not conducted [45].Narrative synthesis was performed when meta-analysis was not possible due to too few studies reporting outcome data similar enough to pool, or when it was not possible to convert data reported to ORs [46].Data were tabulated and grouped into thematic categories relating to maternal and infant health outcomes (a priori categories).Further data-driven thematic categories were developed based on the specific outcomes reported in the included studies (e.g., birth weight) and how these data were presented (e.g., birth weight for gestational age z-score).Narrative summaries accompany the tables to report associations between food insecurity in pregnancy and health outcomes.

Search results
There were 24,223 results from the database searches: 11,377 were duplicates, 12,633 were excluded at the title and abstract screening stage, and 213 full texts were screened (Fig 1).There were 2,422 additional records identified from supplementary searches (543 from websites and 1,879 from forwards and backwards citation chaining).Details of contact with study authors are provided in S3 Table.
While overall the studies reflected general maternity populations with their inclusion criteria, albeit usually singleton pregnancies and >18 years of age, there were some studies that  applied criteria that may have excluded food insecure populations (S5a Table ).These included restrictions to English speaking populations [29,50,53,60,66], exclusion of women in prisons [62,63,68], or First Nations reserves and settlements, Armed Forces, or living in care homes [68].There were various and inconsistent participant characteristics reported by the included studies including mean, median, or categories of maternal age (n = 15); ethnic group or race (n = 25); parity, previous pregnancy, or having other children (n = 15); income brackets, quintiles, poverty line, and income need ratio (n = 15); various categories of educational attainment (n = 21); employment status (n = 9); relationship status (n = 16); and other socioeconomic indicators including measures of social security, WIC use, financial support, and housing tenure, occupancy, and neighbourhoods (n = 16) (see S5b Table ).The mean or median age ranged from 18.6 to 33 years.Ethnic and racial representation included white, black, Hispanic, Latina, Asian, Pacific Islander, Pakistani, Bangladeshi, Indian, American Indian, Alsakan, and Indigenous groups.Majority populations included in the studies were white (n = 10 studies), Hispanic or Latina (n = 5), black (n = 2), South Asian (n = 1), or equal distribution between 2 or more groups (n = 5).Education less than high school equivalent ranged from 8.7% to 34.7% of participants.Women were more likely to have had previous pregnancies and/or other children in 9 studies, and a first pregnancy in 4 studies.There were more studies reporting that their participants were either married or in a relationship (n = 13) than unmarried or single (n = 3).More studies also reported a higher proportion of their participants to be in employment (n = 6) than not employed (n = 3).[49].Three studies reported associations between food insecurity and cesarean delivery, and all were pooled in a meta-analysis (Fig 3).This included 2 high-quality studies [50,51] and 1 medium quality [48] with a pooled sample size of 1,256 women.Pooled analysis showed no association between food insecurity and cesarean section (OR 1. 42 [29,48,57].Results for chronic or gestational hypertension also showed an inconsistent association with food insecurity, with 1 study reporting a positive association (OR 16.94, 95% CI [7.79, 36.85])[48] and 2 others showed no association (OR 1.45, 95% CI [0.65, 3.24] [29] and OR 1.22, 95% CI [0.94, 1.59]) [67].One study reported a combined outcome of hypertension or diabetes which showed no association with marginal or moderate/severe food insecurity compared with food security (ARR 0.99, 95% CI [0.51, 1.92] and 1.44, 95% CI [0.88, 2.35], respectively) [30].

Food insecurity and maternal physical health in pregnancy
One study [65] reported significantly increased odds for a variety of oral health and dental care needs among women experiencing food insecurity, such as needing to see a dentist for a problem, not knowing it was important to care for teeth, and not talking about dental health with providers (AORs ranged from 1.29, 95% CI [1.04, 1.59 to 1.91], 95% CI [1.62, 2.25]).One study reported a positive association between serum perfluorooctane sulfonate (PFOS) Forest plot for meta-analysis of the association between cesarean section and food insecurity.Legend: Categories of food insecurity and food security: adequate food security vs. inadequate food security, Cheu and colleagues, 2020 [50]; food secure vs. food insecure, Cooper and colleagues, 2022 [51]; food insecurity vs. not food insecurity, Bihan and colleagues, 2023 [48].Measurement of FI: Cheu and colleagues, 2020 used USDA HFSSM 10 items [50]; Cooper and colleagues, 2022 used USDA HFSSM 18 items [51]; Bihan and colleagues, 2023 used 1 single item adapted from USDA HFSSM [48]. https://doi.org/10.1371/journal.pmed.1004450.g003

Food insecurity and infant health outcomes
Three studies reported data for birth weight; 2 were high quality [53,55] and 1 medium quality [56] with a pooled sample size of 594 women.Four studies reported data for SGA and LGA, 3 high quality [30,50,67] and 1 medium quality [48] with a pooled sample size of 12,739 women.Pooled results showed no association between food insecurity and mean birth weight .Additional data reported that could not be pooled in the meta-analysis (see S9 Table ) showed no association with food insecurity and birth weight for gestational age z-score [53] or low birth weight (RR 1.21, p = 0.51 and OR 1.39, 95% CI [0.45, 4.29]) [29,50].However, one study reported a positive association between food insecurity and high birth weight in 2 different adjusted models: AOR 2.07 (95% CI [1.07, 3.98]) when adjusting for child sex, race, ethnicity, parent age, education, income, and study site; and AOR 1.96 (95% CI [1.01, 3.82]) when adding poor neighbourhood food environment to the model (S9 Table) [52].

Discussion
This systematic review has extensively explored associations between women experiencing food insecurity in HICs post the 2008 global financial crisis, and pregnancy health outcomes for women and their infants.The meta-analysis identified positive associations between food insecurity and GDM, high stress levels, and mood disorder, but no association with birth weight, SGA, LGA, preterm delivery, NICU admission, or cesarean delivery.Narrative synthesis also identified patterns for associations between food insecurity and some maternal physical health outcomes including dental health, and serum PFOS, but evidence was limited for infant health outcomes.Further, there were highly consistent data for associations between food insecurity and a range of maternal mental health outcomes that could not be included in metaanalysis due to a limited number of studies reporting similar enough data for pooling.These included increased risks for depression, anxiety, and mental health disorders, and women experiencing food insecurity being more likely to also experience stress events, depressive symptoms, poor resilience, and lower perceived mental health.
The associations between food insecurity and maternal mental health outcomes could be a consequence of the compounded effect from intersecting and interconnected inequalities such as gender and class.In previous studies, women have described that due to a lack of income they are unable to participate in normal consumer routes for purchasing food, or participate socially, and describe feelings of tension between their nutritional desires and their inability to meet them in addition to ongoing uncertainty around access to food [23].Pregnancy, birth, and early parenthood represent life-changing moments.Moments of change can instigate a period of poorer mental health or worsen preexisting mental health conditions [71].Therefore, a concern is that women with food insecurity experience more severe mental health challenges (double burden on mental health) as they experience the mental health costs associated with food insecurity in addition to the potential mental health costs from pregnancy.Maternal mental health must be acknowledged, not only to improve maternal health and well-being but also because children exposed to maternal mental ill health are at increased risk of experiencing psychological and developmental disturbances [72,73].The findings from this review make it evident that screening for food insecurity during pregnancy and mental health conditions in HICs is necessary to ensure care for mother and baby.The World Health Organisation (WHO) recognises the importance of screening, diagnosing, and managing perinatal mental health conditions, publishing new guidelines for integrating perinatal mental health into maternal and child health services in late 2022 [71].Specific to poverty, WHO recommend health services monitor infants for malnutrition and forge links with community providers that provide services addressing the wider social determinants of health, such as housing associations.Further, they encourage women to form strong social networks that provide support during financially difficult times, a sentiment that will be difficult to achieve given the challenges of living with poverty, mental health, and pregnancy.
The finding that pregnant women experiencing food insecurity have significantly increased odds of GDM is concerning as developing GDM can result in poorer health outcomes for both mother and baby.These include LGA infants, complications at birth, and increased life course risk of developing T2DM for women and their children [74].Children born to women with GDM could be at risk of metabolic disease later in life [75], thus food insecurity during pregnancy can have important potential life course implications for the child.Risk factors for GDM include maternal overweight and obesity.Women experiencing food insecurity during pregnancy are significantly more likely to be living with obesity [76] which in itself increases the risk of obesity development in the child [77].Therefore, food insecurity compounds the potential increased risks for the child relating to both maternal obesity and GDM.Potential mechanisms underpinning the association between food insecurity and GDM include the wider determinants of maternal obesity such as housing conditions and food environment.Accounts of food insecurity report poor kitchen facilities and utensils to cook and store healthy meals [23].Further, food insecurity is associated with energy-dense diets and living in obesogenic environments, thus increasing the risk of developing obesity and GDM [78,79].Another potential mechanism explaining the association between food insecurity and GDM could be the dysregulated eating patterns characteristic of food insecurity which could alter metabolic regulation contributing to the development of GDM [80].Eating patterns for those experiencing food insecurity are typified by food being consumed when food is available, and restricted when food shortages exist [23].Cycles of food abundance at the beginning, then food shortages at the end of the month have been reported [81].Thus, charitable food aid, social support, or income distributed weekly rather than monthly could help mitigate these eating patterns.A study has also shown food insecurity is indirectly linked with higher body mass index via maladaptive coping mechanisms such as emotional eating [82], which this review has identified could be triggered by significant high levels of stress women experience.Finally, among women, food insecurity is associated with nutritionally poor diets, lacking fruit and vegetables which could deprive food insecure women of the protective nutrients required to prevent disease such as GDM [83][84][85].
This review identified limited data for infant health outcomes, and data that were included showed a lack of association between food insecurity and SGA, LGA, preterm delivery, and NICU admission.Some of these data are in conflict with evidence from LMICs showing that low birth weight was higher in food insecure groups compared to food secure groups [86][87][88], potentially due to differences in context between LMICs and HICs relating to food insecurity and availability of types of foods.However, our meta-analysis was limited by the limited number of studies reporting infant outcomes and the context being in North America, which may have impacted on the lack of significance in pooled results despite an overall pattern towards positive directions of effect.The USA context may bias the results due to a potential intervention effect from the well-established Special Supplemental Nutrition Program (SNAP) for Women, Infants and Children (WIC) in the USA [89].Evidence shows that WIC reduces food insecurity and can improve pregnancy outcomes such as gestational hypertension [90], preterm delivery [91], SGA and NICU admission [92,93].However, other HICs do not have embedded support in place like WIC, thus further research in other contexts is needed to fully understand the impact of food insecurity on pregnancy outcomes.

Strengths and limitations
There are several strengths to this systematic review including following gold standard methods such as registering the study protocol in advance of conducting the review on PROSPERO, using a rigorous multistranded search strategy to supplement our database searches with forwards and backwards citation chaining, independent duplicate screening, data extraction and quality assessments, and contacting study authors to overcome potential publication bias.As our review was holistic in terms of pregnancy outcomes, we did not include any specific outcome terms in our search strategy as this may have omitted studies reporting relevant maternal and infant pregnancy outcome data that were not in a predefined list of outcome search terms.Further, the use of the PRISMA and MOOSE guidelines helps to improve the transparency and scientific merit when reporting systematic reviews and meta-analysis [32,94,95].While we were not able to conduct statistical analysis for publication bias as we had originally planned due to insufficient studies pooled in each meta-analysis, there were multiple included studies reporting nonsignificant results.This suggests that the potential for publication bias relating to nonsignificant results was minimal.
There are also several limitations to the review.We identified a small number of studies reporting consistent outcomes or reporting data in a way that could be pooled, thus limiting our ability to conduct meta-analysis.We made an a priori decision to conduct meta-analysis if there were 3 or more studies reporting similar enough exposures of food insecurity and pregnancy outcomes, resulting in 9 meta-analyses reported in this review.However, there were wide variations in the sample sizes of included studies overall and in the pooled sample sizes for the meta-analysis.These ranged from 594 women for birth weight meta-analysis to 19,440 women for high stress meta-analysis; this potentially contributed to some of the heterogeneity observed in analysis.Additionally, as sample size influences the weighting of any individual study, the inclusion of one large study with multiple smaller studies will impact on the weighting towards a single study.Understanding the sources of heterogeneity was particularly relevant for maternal high stress.Although we had planned to explore sources of heterogeneity using meta-regression methods, this was not possible as there were fewer than 10 studies included.While the overall effect size of this meta-analysis should be interpreted with caution due to the potential influence of one study and wide confidence intervals, there was a consistent positive association between food insecurity and stress in all studies reporting these outcomes.The narrative synthesis also identified maternal mental health to be consistently significantly associated outcome with food insecurity across all papers reporting these related outcomes.The ability to conduct more robust meta-analysis for maternal mental health would enable us to draw more robust and generalisable conclusions about the extent that food insecurity is associated with maternal mental health and explore contributing factors and sources of heterogeneity between studies.Another source of heterogeneity limiting this review was the use of different food insecurity measures between studies.This meant we were unable to compare the different food insecurity levels between studies, thus, impacting the interpretation of the association of food insecurity on different health outcomes across included studies.Further, meta-regression was not possible; we were therefore unable to explore the influence of studies adjusting for confounding variables.
Of the included studies none were classified as low quality, and the majority of studies contributing data to meta-analysis were high quality.All studies included in meta-analysis for stress, NICU, mood disorders were high quality, with 1 or 2 medium quality studies contributing to GDM, LGA, SGA, birth weight, cesarean, and preterm delivery meta-analysis.A low response from study authors upon requesting additional information, such as frequency data to enable calculation of ORs and 95% CIs, impacted the meta-analysis as some studies were not able to be included.Further, although there were some studies that reported adjusted models for associations between food insecurity and pregnancy outcomes, there was inconsistency in the factors included in the adjusted models.Many studies did not include adjustments at all, and there may be issues with residual confounding for unmeasured, or poorly measured variables.While it would be unethical to control for this in studies of food insecurity, for example, by using randomisation methods, thorough consideration must be given to residual confounding given the intersecting nature of inequalities and the likelihood that food insecurity does not exist in isolation of other factors that may also be influencing the pregnancy outcomes.
The use of different tools to assess food insecurity and the use of different food insecurity categories, most notably in classification of persons between the marginal or moderate versus low or very low level of food security presents a limitation to this review as it introduces study heterogeneity.The majority of included studies conducted in North America used the validated USDA HSFFM or its short/modified version.The USDA HFSSM was developed in the USA and has been used to monitor prevalence of food insecurity annually since 1995 [96].The use of its different forms (short forms or single item) may introduce classification inconsistencies as the use of a short form, or a single item, is not as rigorous a measurement of the severity of household food insecurity.For example, in Australia, a single-item measure of food insecurity which was incorporated in their National Health Survey only tends to capture marginal food insecurity [97].In addition, the prevalence of food insecurity is commonly estimated on the basis of the number of "food insecure" conditions reported (in the form of affirmative responses) on a food insecurity scale [98].While the USA and Canada share the same food insecurity measurement, the HFSSM, they use different classification system for different level of food insecurity severity [99].The USA requires 3 affirmative responses before classifying a household as food insecure, "food secure" is classified if there is no affirmative response or up to 2 affirmative responses to questions on the HFSSM, "low food security" if there are 3 to 7 and "very low food security" if there are 8 or more [100].Whereas, in Canada, one affirmative response on either the adult or the child scale is taken as indicating food insecurity (marginal food security level) [100][101][102].Thus, when applied to the same data, the USA's classification system would yield a lower figure for food insecurity and may have less sensitivity to detect vulnerability to negative health consequences associated with food insecurity than the Canadian system [101,103].Therefore, there is a need for more consistent measures to facilitate between study comparisons.In addition, given the impact of food insecurity on maternal physical and mental health outcomes, it is important to use the full-scale measure of food insecurity and consider the marginal level of food insecurity to better understand the relationship between food insecurity and health outcomes.
Finally, almost all the studies included in this review were conducted in North America, primarily the USA (n = 20), with 3 from Canada and 2 from Europe (1 France, 1 United Kingdom).This highlights the gap in the literature investigating the impact of food insecurity in pregnancy across HICs.Unlike in North America where the food aid system is institutionalised, and research has been conducted for many years, in European HICs it was not until after the global financial crisis of 2008 and the parallel rise of charitable food aid that food insecurity has become an urgent issue, with increasing prevalence widely impacting on health status and systems such as maternity.This gap also limits the generalisability of this review's results to other HIC contexts.Importantly, there may be differences in the impact of food insecurity in pregnancy in other HIC contexts where there is a high prevalence of food insecurity and different, or absent, support mechanisms embedded within public health and maternity services.

Conclusions
This review has identified that food insecurity in pregnancy is significantly associated with some poorer maternal health, particularly relating to GDM and mental health.However, little data were available outside of the North America context and thus a key recommendation from this review is for further research in wider HIC contexts to fully explore potential associations, especially in countries with high prevalence of food insecurity.Findings of this review are important to inform the improvements in maternity services to better be able to support women experiencing food insecurity, particularly relating to a need for screening for food insecurity in conjunction with GDM and mental health.Importantly, these results identify potential life course implications of experiencing food insecurity in pregnancy for both women's health and their children's health, alongside increasing prevalence of food insecurity in many HICs.This requires urgent public health interventions to tackle poverty and evaluation should factor in the potential benefits of alleviating food insecurity in pregnancy on the shortand long-term health and well-being of women and their children.