Global prevalence of underweight and overweight/obesity among migrant workers: a systematic review and meta-analysis Authors: Herlin Ajeng Nurrahma, Hidayat Arifin, Anung Ahadi Pradana, Herry Susanto, Chiu-Li Yeh, Ya-Ling Chen, Hitoshi Shirakawa, Kuei-Ru Chou, Suh-Ching Yang Published: 2026 ABSTRACT BackgroundMigrant workers are increasingly recognised as a nutritionally vulnerable population due to unstable working conditions, disrupted food environments, and limited access to health services. However, global estimates of malnutrition in this group remain fragmented. We aimed to identify the global prevalence of underweight and overweight/obesity among migrant workers and identify key moderating factors.MethodsWe conducted a systematic review and meta-analysis following PRISMA guidelines. We searched seven databases without restrictions on language, region, or year of publication. Observational studies reporting the prevalence of underweight (body mass index (BMI) < 18.5 kg/m2) or overweight/obesity (BMI ≥ 25 kg/m2) among migrant workers were included. We calculated pooled prevalence estimates using generalised linear mixed models with random effects. We assessed heterogeneity with I2 and Cochran’s Q statistics. Moderator analyses were conducted using meta-regression and subgroup comparisons. Further, we appraised study quality using the Joanna Briggs Institute tools, and assessed the certainty of the evidence using GRADE.ResultsWe included 30 studies involving 135 404 migrant workers with a mean age of 32.3 years (standard deviation = 11.1). The pooled prevalence was 4.7% (95% confidence interval (CI) = 3.1, 6.9) for underweight and 43.7% (95% CI = 39.8, 47.6) for overweight/obesity, suggesting a substantial double burden of malnutrition. Underweight was more common among local migrants and in lower-middle-income settings, particularly in South Asia, whereas overweight/obesity was more common among international migrants, especially in high-income countries and in Europe and Central Asia. Moderator analyses identified female proportion, destination country or region, income level, and marital status as significant moderators for both outcomes.ConclusionsMigrant workers experience a substantial double burden of malnutrition, although the pattern varies across migration contexts. Undernutrition remains a concern in lower-income settings, whereas overweight and obesity are more common in high-income settings. Integrating nutrition-sensitive strategies into labour and migration health policies, including healthier food environments, routine nutritional screening, and culturally appropriate interventions, may help reduce health inequalities in this population.RegistrationINPLASY: 202540030. Urbanisation and globalisation have accelerated labour migration from developing countries to urban areas within national borders and to high-income countries, as individuals seek better employment and living conditions [1,2]. These global movements have resulted in growing populations of migrant workers, with over 169 million people engaged in international labour migration worldwide [1]. A migrant worker is broadly defined as an individual who relocates across geographic boundaries to engage in paid employment, either within a country (internal migration) or across international borders (international migration) [1,3]. However, definitions of migrant workers vary across disciplines and institutions, leading to inconsistencies in population scope and limiting comparability across studies. Given the heterogeneity of migrant populations, a structured classification framework was applied in this review to ensure conceptual consistency across studies. In this study, migrant workers were defined according to terminology from the International Organization for Migration and the International Labour Organization, encompassing individuals engaged in economically motivated labour migration, including internal rural – urban migrants, international migrant workers, seasonal migrant workers, and temporary contract migrant workers [1,4]. During migration, individuals frequently experience lifestyle and dietary disruptions due to unfamiliar food environments, limited financial resources, and demanding work conditions [5,6]. These stressors are compounded by legal uncertainty, social marginalisation, and limited access to healthcare in both transit and host countries [7], making migrant workers especially vulnerable to malnutrition. Malnutrition, defined as a condition resulting from an imbalance, deficiency, or excess of nutrients that negatively affects body composition and health outcomes [8], has increasingly been recognised among migrant populations. Its determinants are multifactorial, including food insecurity, limited social integration, low dietary diversity, age-related vulnerabilities, and precarious living or working conditions [5,9,10]. This can result in undernutrition, micronutrient deficiencies, and growth-related issues such as stunting and wasting [7,8,11]. At the same time, migrants in high-income countries frequently experience overweight and obesity alongside ‘hidden hunger’ characterised by energy-dense but nutrient-poor diets, compounded by limited food autonomy and financial constraints [9,12]. These nutritional challenges not only compromise physical health but also reduce work productivity, diminish quality of life, and contribute to long-term healthcare burdens [13–16]. From a public health nutrition perspective, malnutrition among migrant workers represents a systemic failure in ensuring equitable access to nutrition and health services for vulnerable populations. Addressing this issue requires integration of nutrition into broader public health and migration policies through initiatives such as nutritional surveillance, workplace health programmes, culturally tailored dietary education, and food environment reform [1]. Without interventions, malnutrition among migrant workers will continue to fuel preventable diseases, widen health disparities, and increase economic burdens on health systems in both sending and receiving countries [5]. Previous studies reported various forms of malnutrition among migrants across regions. For instance, undernutrition remains a major concern in poor households with mothers who are migrant workers [7], while overweight and obesity are increasingly observed among migrants in high-income countries and among rural-to-urban migrants [17–19]. A systematic review by Ankomah et al. reported that overweight/obesity among migrants and refugees ranged from 11.1–42.0%, while undernutrition ranged from 0.3–17%, highlighting the double burden of malnutrition in these populations [9]. However, the existing evidence base remains fragmented and heterogeneous, with most studies focusing on specific subpopulations (e.g. refugees, women, or children) or limited geographic regions. Importantly, migrant workers, who constitute a large and distinct population with unique occupational, social, and environmental exposures, have not been comprehensively examined globally. Moreover, the influence of migration-related characteristics, such as type of migration (local vs. international) and length of stay, on nutritional outcomes remains poorly understood. Prolonged duration of stay may exacerbate nutritional vulnerability through persistent socioeconomic disadvantage or, alternatively, promote acculturation to obesogenic food environments, yet these dynamics have rarely been systematically assessed. In addition, variability in study design and methodological quality further complicates the interpretation of existing findings. Addressing these gaps, the present study conducted a global meta-analysis to estimate the pooled prevalence of underweight and overweight/obesity among local and international migrant workers and to examine key moderating factors influencing malnutrition. METHODS Search strategy and data sources In this study, we followed guidelines established by the Meta-analysis of Observational Studies in Epidemiology and the updated PRRISMA guidelines [20,21] (Appendix S1 and S2 in the Online Supplementary Document). Our study was registered with the International Platform of Registered Systematic Review and Meta-analysis Protocols (202540030) [22]. Because we used aggregated, anonymised data from publicly available published sources, ethical approval was not required. We conducted an updated comprehensive literature search on 4 March 2026 with the assistance of a university librarian across seven electronic databases – CENTRAL, Embase, PubMed, Scopus, Web of Science, Latin American and Caribbean Literature on Health Sciences, and African Index Medicus. We performed the search without restrictions on country, language, or publication year. To ensure comprehensive coverage, backward citation tracking was performed by screening the reference lists of included studies, and forward citation tracking was used to identify studies that cited them. Additionally, we searched Google Scholar and grey literature sources to identify potentially relevant studies that may not have been captured through database searches. Our search strategy involved a combination of Boolean operators and specialised terms tailored to each database. In CENTRAL, PubMed, Scopus, and Web of Science, we employed both free-text keywords and relevant MeSH terms to refine the search. For Embase, we used Emtree terms, the database’s controlled vocabulary, to enhance search specificity and ensure the inclusion of all criteria. This approach optimised the retrieval of relevant studies by adapting to each database’s unique indexing system. The primary search concepts were structured around three main themes: epidemiological measures (e.g. prevalence and incidence); nutritional condition (e.g. malnutrition, thinness, and overweight); and mobile populations (e.g. migrants, nomads, and transients) (Appendices S3 and S4 in the Online Supplementary Document. Eligibility criteria Regarding the inclusion criteria, the study population had to be comprised of migrant workers, defined according to terminology from the International Organization for Migration and the International Labour Organization, including internal rural-urban migrants, international migrant workers, seasonal migrant workers, and temporary contract migrant workers. These groups were defined as individuals engaged in economically motivated labour migration within or across national borders (Appendix S5 in the Online Supplementary Document) [1,4]. Malnutrition had to be assessed in accordance with World Health Organization (WHO) definitions [8], with being underweight defined as a body-mass index (BMI)<18.5 kg/m2 and overweight/obesity as a BMI>25 kg/m2. Studies had to report quantitative prevalence data on malnutrition among migrant workers and use observational designs, including cross-sectional and cohort studies, which are well-suited for estimating prevalence and identifying trends. Intervention studies (e.g. randomised controlled trials) were also included if baseline nutritional data were available prior to the intervention. We excluded studies that did not focus on malnutrition or migrant workers, involved populations outside the scope of this review, used qualitative or other designs unsuitable for prevalence estimation, or lacked sufficient data for analysis. Screening, data extraction, quality assessment and certainty of evidence assessment Two researchers (HAN and AAP) independently reviewed the data using EndNote, version 21 (Clarivate Analytics, Philadelphia, Pennsylvania, USA). Initially, duplicates were removed, and titles and abstracts of the remaining were screened. This was followed by a full-text review to assess whether the studies met the inclusion criteria. Disagreements were resolved through discussion with a third researcher (SCY.) if necessary. To ensure accuracy and thoroughness, the reviewers contacted the authors of published articles for additional data or clarification. Once eligible full texts were obtained, data extraction was performed by two researchers (HA and HS) who had previously agreed on the information to be extracted. The extracted data included: study characteristics, such as authors, publication year, country, and study design; demographic data and risk factors, such as sample size, gender, age, comorbidities, behavioural risk factors, educational level, migrant setting, and migrant type; and screening tools and malnutrition (underweight and overweight/obesity) prevalence. When a study reported prevalence estimates for multiple independent populations (e.g. different countries, migrant types, or survey years), we treated these estimates as separate data points in the meta-analysis because they represented distinct population samples [23,24]. We appraised the study quality using the Joanna Briggs Institute tool for prevalence studies [23,25], which comprises nine domains that evaluate sampling, measurement, and analysis. Each domain was scored as ‘yes,’ ‘no,’ ‘unclear,’ or ‘not applicable,’ and overall quality was categorised as high (>70%), moderate (50–70%), or low (<50%) based on the number of ‘yes’ scores [26]. We assessed the certainty of evidence using the GRADE approach [27], considering risk of bias, inconsistency, indirectness, imprecision, and publication bias. Ratings were assigned as high, moderate, low, or very low. Downgrading occurred due to serious concerns, such as high heterogeneity (I2 > 95) or publication bias (e.g. funnel plot asymmetry, Peters’ test P < 0.05) [28]. All disagreements during quality or certainty assessments were resolved by consensus or third-party adjudication. Statistical analysis We calculated pooled prevalence estimates of underweight and overweight/obesity using a random-effects meta-analysis, accounting for between-study heterogeneity. Prevalence was defined as the proportion of malnutrition cases among migrant workers in each study and was expressed as a percentage, with corresponding 95% confidence intervals (CIs) and 95% prediction intervals. We assessed statistical heterogeneity using Cochran’s Q test and the I2 statistic. Heterogeneity was considered substantial when I2 exceeded 30%, and the Q test was statistically significant (P < 0.10) [24]. Given the presence of substantial heterogeneity, moderator analyses were conducted to explore potential sources of variability across studies. We performed a random-effects meta-regression using study-level covariates, including mean age (in years), mean duration of stay (in years), and percentage of female participants. Meta-regression coefficients represent absolute percentage-point changes in prevalence associated with a one-unit increase in the moderator variable (e.g. years for mean age or percentage points for female proportion). We modelled prevalence estimates using generalised linear mixed models with random effects. Continuous moderators were entered in their original measurement units to maintain interpretability of the regression coefficients [23,24]. In addition, we conducted subgroup analyses according to study characteristics (study design, publication year, study setting, sample size category, study quality, destination region, and country income level) and participant characteristics (marital status, educational level, migration setting, migration scope, and migrant type). The choice of these moderators was guided by the Social Determinants of Health (SDOH) framework [29]. In this framework, destination region and country income level represent structural determinants that capture broader socioeconomic and policy contexts shaping migrant health, whereas individual characteristics such as education, age, sex, and marital status reflect socioeconomic position. Study setting and migration context, including urban or rural environments, were treated as proxies for intermediary determinants related to living conditions, access to services, and health system context. Using the SDOH framework provided a theory-driven basis for subgroup stratification and strengthened the interpretability of subgroup findings. A P-value <0.05 was considered statistically significant. We conducted sensitivity analyses to evaluate the robustness of the pooled estimates. Specifically, we applied a leave-one-out approach, sequentially excluding the study with the smallest and the largest statistical weight to assess the influence of individual studies on the overall results. Further sensitivity analyses were also conducted according to study design and migrant worker type to examine the consistency of the findings across methodological and population subgroups. We assessed publication bias through visual inspection of funnel plots and Peters’ regression test, with P ≥ 0.10 indicating no statistically significant evidence of small-study effects or publication bias [30]. When asymmetry was observed, or the Peters’ test indicated potential bias, the trim-and-fill method was applied to estimate the number of potentially missing studies and to evaluate the impact of such bias on the pooled estimates [31]. The adjusted pooled prevalence obtained after imputing missing data from potentially missing studies was compared with the original estimate to assess whether publication bias materially influenced the results. Peters’ test suggested possible small-study effects; however, this finding should be interpreted cautiously because funnel plot asymmetry and regression-based tests in prevalence meta-analyses may be influenced by substantial heterogeneity and sample size variation. We performed all analyses using Comprehensive Meta-Analysis, version 3.0 (Biostat, Englewood, New Jersey, USA) [30]. RESULTS Study selection and characteristics We initially identified 10 811 relevant studies from electronic databases. After removing 2827 duplicates, 7984 studies remained. Of these screened by title and abstract, 7824 records were excluded due to irrelevant topics, populations, or study designs. Subsequently, we assessed 160 full-text studies for eligibility; 130 were excluded due to insufficient data or inappropriate measurement tools. As a result, 28 eligible studies were included in the analysis. To ensure the comprehensiveness of the search, we performed an additional manual search on Google Scholar and conducted citation tracking, which yielded three additional eligible studies. Therefore, 30 studies were included in the meta-analysis. From these studies, 22 data points on underweight and 48 data points on overweight/obesity were extracted. Specifically, one study [32] provided malnutrition data by the destination country, one study [33] provided data by migrant type, two studies [34,35] by year of data collection, two studies [6,36] by host country, and two studies [37,38] by category of migrant worker (Figure 1, Table 1). PRISMA flow diagram of article selection process. Table 1Characteristics of included studiesStudyCountryStudy designDemographic data and risk factorsTools, underweight prevalence, overweight/obesity prevalence Sample size and gender Age in years, x̄ (SD); comorbidities; behaviour risk Education* Migrant type and setting Abasilim et al., 2024 [10]USACross-sectionaln = 102; female n = 101 (99%), male n = 1 (1%)NA97) and evidence of publication bias, while no serious concerns were identified for indirectness or imprecision. Overall, these findings indicate substantial variability in prevalence estimates across studies and highlight the need for more standardised and methodologically robust research on migrant nutrition (Appendix S8 in the Online Supplementary Document). Results of the meta-analysis and sensitivity analysis Pooled prevalence of underweight The pooled prevalence of underweight among migrant workers was 4.7% (95% CI = 3.1, 6.9) (n = 26 244) [6,11,14,34,36,38–40,42–44,46,47,49,52,56,57]; the corresponding prediction interval value ranged between 0.7 % and 33.6 % for potential future studies. Heterogeneity was considerable (Q = 835 442.19, P < 0.001, I2 = 97.5), and a random-effects model was used (Figure 2, Panel A). Subgroup analysis revealed a higher prevalence of underweight among local migrants (12.5%; 95% CI = 10.4, 15.1) than among international migrant workers (5.0%; 95% CI = 3.4, 7.2) (Figure 3, Panels A and B). Visual inspection of the funnel plot and Peters’ regression (P = 0.02) suggested asymmetry and possible small-study effects. However, Duval and Tweedie’s trim-and-fill method imputed zero missing studies, so the adjusted pooled estimate was essentially unchanged (no meaningful adjustment) (Appendix S9 and S10 in the Online Supplementary Document). Pooled prevalence among migrant workers. Panel A. Underweight. Panel B. Overweight/obesity among migrant workers. Pooled prevalence. Panel A. Underweight among local migrant workers. Panel B. Overweight/obesity among local migrant workers. Panel C. Underweight among international migrant workers. Panel D. Overweight/obesity among international migrant workers. Pooled prevalence of overweight/obesity The pooled prevalence of overweight/obesity among migrant workers was 43.7% (95% CI = 39.8, 47.6) (n = 135 404) [6,10,11,14,15,32–53,55,56]. The corresponding prediction interval value ranged between 20.6% and 69.8% for potential future studies. Heterogeneity was high (Q = 75 995.83, P < 0.001, I2 = 99.4), and a random-effects model was therefore applied (Figure 2, Panel B). By migrant type, international migrants had the highest prevalence (50.5%; 95% CI = 43.3, 57.7), followed by local migrants (33.4%; 95% CI = 31.0, 35.8 (Figure 3, Panels C and D). Regarding publication bias, the funnel plot demonstrated asymmetry. Publication bias was suggested by funnel plot asymmetry and supported by Peters’ regression test (P = 0.01). Trim-and-fill analysis indicated potential small-study effects, with five studies estimated to be missing on the right side of the funnel plot. After imputing these studies using a random-effects model, the pooled prevalence increased slightly from 43.7% (95% CI = 39.8, 47.6) to 46.7% (95% CI = 42.5, 51.0) (Appendix S1 and S12 in the Online Supplementary Document). The small difference between the observed and adjusted estimates indicates that the pooled prevalence remained stable, suggesting that potential publication bias had limited influence on the overall findings. Sensitivity analysis Sensitivity analyses supported the robustness of the pooled estimates. For underweight, excluding the study with the smallest statistical weight [34] yielded a revised prevalence of 5.1% (95% CI = 3.5, 7.4), whereas excluding the study with the largest statistical weight [56] resulted in a prevalence of 4.3% (95% CI = 2.9, 6.5). Additional leave-one-out analyses produced similar estimates. When we examined studies by migrant worker categories, the pooled prevalence across all categories was 4.8% (95% CI = 3.2, 7.3). Excluding a study representing local migrants [38,49,56] yielded a comparable estimate of 5.1% (95% CI = 3.7, 7.0), whereas exclusion of a highly weighted study representing international migrant workers [6,14,36,38,40,42–44,46,47,52,54,57] resulted in a prevalence of 4.7% (95% CI = 3.1, 7.1). For overweight/obesity, excluding the study with the smallest statistical weight [56] yielded an estimated prevalence of 44.2% (95% CI = 40.3, 48.2), whereas excluding the study with the largest statistical weight [32] resulted in a prevalence of 42.7% (95% CI = 38.9, 46.5). When examined across migrant worker categories, the pooled prevalence was 44.1% (95% CI = 40.5, 47.8). Excluding a study of short-term migrants [37] produced a comparable estimate of 44.2% (95% CI = 40.6, 47.9), while excluding a highly weighted study of international migrant workers [6,10,14,32,36,38,40–44,46–48,50–52,54,57] yielded a slightly lower prevalence of 40.6% (95% CI = 36.6, 44.8). Additional sensitivity analyses that excluded the cohort study [52] yielded a similar estimate of 43.9% (95% CI = 39.9, 47.9). Overall, these analyses indicate that the pooled estimates were stable and not materially influenced by individual studies. Result of the moderator analyses Underweight among migrant workers Meta-regression and subgroup analyses were performed to identify potential sources of heterogeneity among the included studies. The meta-regression analysis revealed female proportion (6.6%; 95% CI = 5.4, 7.7) was a significant moderator of the pooled prevalence of being underweight among migrant workers, even though the mean age (2.2%; 95% CI = –3.9, 8.3) and length of stay (3.3%; 95% CI = –1.3, 8.3) non-significantly moderated it (Table 2). Table 2Result of moderator analysisVariablesUnderweightOverweight/obesity Study size/participants, n Pooled estimate, % (95% CI) P-value Study size/participants, n Pooled estimate, % (95% CI) P-value Meta regression analysis Mean age 16/26 2442.2 (–3.9, 8.3)0.48040/118 7565.1 (3.2, 7.0)<0.001 Length of stay 9/91843.3 (–1.9, 8.5)0.21514/60 620–0.1 (–3.1, 2.9)0.0951 Female 16/79876.6 (5.4, 7.7)<0.00141/54 0414.8 (4.5, 5.2)<0.001 Subgroup analysis Study characteristics Cohort study 1/767.9 (3.6, 16.5)<0.0011/7632.9 (23.3, 44.2)<0.001 Cross-sectional study 21/26 1284.6 (3.0, 6.9) 47/135 32843.9 (39.9, 47.9) Publish year 2011–2015 NANA<0.00113/100 48632.7 (30.7, 34.7)<0.001 2016–2020 10/12 3095.7 (3.9, 8.4) 18/19 82550.1 (40.7, 59.5) 2021–2025 11/13 9353.8 (2.0, 7.3) 15/15 09344.7 (36.7, 52.9) Setting Community 17/15 4096.0 (4.4, 8.2)<0.00142/124 27842.3 (39.1, 45.5)<0.001 Hospital 5/10 9762.0 (1.0, 4.0) 6/11 12653.7 (42.5, 64.5) Sample size category <1000 15/11 5616.5 (4.8, 8.9)<0.00132/11 65144.7 (37.4, 52.3)0.003 >1000 7/14 5932.5 (1.0, 6.0) 16/123 75341.5 (35.8, 47.5) Study quality* High 17/16 4074.0 (2.5, 6.5)<0.00141/126 30843.4 (39.2, 47.7)0.002 Moderate 4/97617.6(3.2, 16.9) 5/521244.6 (30.5, 59.6) Low 1/767.9 (3.6, 16.5) 2/388446.5 (22.5, 72.2) Destination country by continent East Asia and the Pacific 10/11 4834.5 (2.7, 7.3)<0.00122/105 51136.6 (33.7, 39.7)<0.001 Europe and Central Asia 1/4507.3(5.3, 10.1) 6/346365.2 (45.8, 80.6) Latin America and the Caribbean 4/9387.6 (4.3, 13.3) 5/108448.9 (35.7, 62.1) Middle East and North Africa 4/11 5491.3 (0.7, 2.4) 4/11 54961.1 (48.9, 72.1) North America NANA 6/419840.0 (25.1, 56.9) South Asia 3/813112.5 (10.4, 15.1) 3/813134.8 (30.2, 39.8) Sub-Saharan Africa NANA 2/145942.8 (11.8, 80.7) Destination country by income level High 8/14 7211.8 (1.2, 2.6)<0.00122/27 41353.2 (43.9, 62.2)<0.001 Upper middle 11/33927.3 (5.6, 9.5) 21/98 40135.7 (33.9, 37.5) Lower middle 3/813112.5(10.4, 15.1) 5/959035.8 (26.6, 46.2) Sample characteristics Marital status Single 10/533225.1 (19.1, 32.4)<0.00115/16 75422.8 (17.3, 29.6)<0.001 Married 10/533265.5 (49.1, 78.9) 15/16 75469.0 (57.3, 78.7) Widow/widower 4/229624.4 (16.8, 34.0) 5/110919.1 (12.1, 28.7) Education level† Lower than secondary level 12/769731.3 (18.5, 47.9)0.86834/117 31449.2 (44.5, 53.9)0.761 More than equal secondary level 12/769773.5 (56.4, 85.6) 34/117 31451.8 (47.1, 56.5) Migrant setting Rural 2/72850.8 (0.6, 1.0)<0.00112/15 00640.4 (28.6, 53.4)0.011 Urban 19/15 4855.8 (4.1, 8.1)35/120 39845.3 (41.2, 49.5)CI – confidence interval, NA – not applicable*The Joanna Briggs Institute (JBI) quality assessment tool was used to evaluate the quality of the included studies, with the total score divided into three categories: high quality, with scores of >70.0%; moderate quality, with scores 50.0–70.0%; low Quality, with scores <50.0%.†Education level:70.0%; moderate quality, with scores 50.0–70.0%; low Quality, with scores <50.0%. †Education level: