Prevalence of normal weight obesity among adults in Southeast Asia: Insights from a systematic review and meta-analysis

Abstract

Background

Normal Weight Obesity (NWO) describes individuals with a normal body mass index (BMI) but elevated body fat percentage (BF%), placing them at increased risk for cardiometabolic complications. This condition is particularly relevant in Southeast Asian populations, where visceral adiposity occurs at lower BMI thresholds. However, regional pooled prevalence data are limited.

Methods

A systematic search was conducted in PubMed, EMBASE, Scopus, Web of Science, and Google Scholar for studies published between January 2010 and May 2025. Eligible studies included observational data on adults (≥18 years) from Southeast Asia reporting NWO prevalence, defined by normal BMI (18.5–24.9 kg/m²) with validated surrogate indices indicative of excess adiposity. Two reviewers independently conducted screening, data extraction, and quality appraisal using the Joanna Briggs Institute (JBI) tool. A random-effects meta-analysis was performed using the Freeman-Tukey double arcsine transformation, with heterogeneity assessed via the I² statistic.

Results

Eight studies involving 9028 participants from five Southeast Asian countries (Philippines, Malaysia, Thailand, Myanmar, Singapore) were included. The pooled prevalence of NWO was 57 % (95 % CI: 37 %–75 %), with study-specific estimates ranging from 22.8 % to 82 %. Heterogeneity was high (I² = 98.9 %). Most studies used bioelectrical impedance analysis (BIA) for body fat assessment and were rated as moderate to high quality.

Conclusion

NWO is common among adults in Southeast Asia, especially in women and young adults. These findings highlight the limitations of BMI as a screening tool and support the integration of body fat assessments into public health screening and clinical protocols for early risk detection.

1. Introduction

Obesity continues to pose a major public health burden, contributing to increased morbidity and mortality through its association with non-communicable diseases (NCDs) such as diabetes, cardiovascular diseases, and certain cancers [1]. Conventionally, obesity is defined using body mass index (BMI), with a cutoff of ≥30 kg/m², serving as a proxy for excess body fat [2]. However, BMI does not accurately reflect body composition or fat distribution, particularly visceral fat, which plays a more critical role in metabolic risk [3].

Emerging evidence suggests that individuals with normal BMI (18.5–24.9 kg/m²) may still exhibit elevated body BF% and be at risk of metabolic disorders such as insulin resistance, hypertension, and dyslipidemia [4]. This phenotype—termed Normal Weight Obesity (NWO)—is characterized by a normal BMI but an abnormally high BF%, commonly >30 % [5,6]. Individuals with NWO may remain clinically undiagnosed due to their seemingly healthy weight, thereby delaying necessary intervention [7].

Several studies from Asian populations, particularly in Singapore and Vietnam, have highlighted a high prevalence of NWO and associated cardiometabolic risks among adults with normal BMI [[8], [9], [10]]. Asian populations, due to their unique body fat distribution, tend to have higher fat percentages at lower BMI levels compared to Western counterparts, justifying lower BMI cut-offs in these regions [3,9]. Consequently, the conventional BMI metric underestimates true adiposity and risk in Southeast Asian populations [3,10].

With increasing urbanization, sedentary behaviour, and dietary changes across Southeast Asia, the prevalence of NWO is expected to rise. However, due to their BMI-defined “normal” status, such individuals often escape routine screening and early preventive efforts [11,12]. Despite growing national studies on NWO from countries such as Thailand, Singapore, Indonesia, and Vietnam, there is currently no regional synthesis estimating the pooled prevalence.

This systematic review and meta-analysis aim to estimate the pooled prevalence of NWO among adults in Southeast Asia and to examine how prevalence varies across measurement methods, age groups, sexes, and countries.

The working hypothesis is that the prevalence of NWO in Southeast Asian populations is substantially high and variation in prevalence is influenced by demographic and methodological factors.

2. Methods

2.1. Study design and reporting framework

This systematic review and meta-analysis were conducted in accordance with the PRISMA 2020 (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines to ensure methodological transparency and consistency in reporting [13]. The review protocol was registered with the International Prospective Register of Systematic Reviews (PROSPERO) under registration number CRD420251168386. The review aimed to synthesize observational studies that reported the prevalence of NWO among adults in Southeast Asia, drawing from literature published between January 1, 2010, and May 31, 2025.

To define the research question and eligibility boundaries systematically, a PICOS framework was used as summarized in Table 1.

Table 1.

PICOS framework defining eligibility criteria for the systematic review on NWO in Southeast Asia.

Component	Description
Population (P)	Adults aged ≥18 years residing in Southeast Asian countries (Brunei, Cambodia, Indonesia, Laos, Malaysia, Myanmar, Philippines, Singapore, Thailand, Timor-Leste, Vietnam) with normal BMI (18.5–24.9 kg/m2, using WHO/Asian cut-offs).
Intervention/Exposure (I/E)	Presence of NWO defined primarily as normal BMI with elevated body-fat percentage (≥30 % for women, ≥25 % for men) (as per De Lorenzo et al., 2006 (14) and WHO Expert Consultation, 2004 (15) assessed using validated techniques such as Bioelectrical Impedance Analysis (BIA), Dual-Energy X-ray Absorptiometry (DXA), or standardized skinfold methods. Studies using validated surrogate measures (e.g., waist circumference ≥90 cm in men/≥ 80 cm in women, waist–hip ratio ≥0.90 M/≥ 0.85 F, or presence of ≥3 metabolic abnormalities per NCEP/ATP III criteria) were also included when direct BF% data were unavailable.
Comparator (C)	N/A
Outcomes (O)	Primary outcome: Prevalence or proportion of NWO among adults with normal BMI. Secondary outcomes: Subgroup prevalence by sex, age, or method of body-fat assessment; association with surrogate metabolic indicators.
Study Design (S)	Observational (cross-sectional or cohort) population-based studies reporting NWO prevalence or proportion.

2.2. Eligibility criteria

We included population-based, cross-sectional, or cohort studies reporting the prevalence of NWO among adults (aged 18 years and above) residing in Southeast Asia, defined as Brunei, Cambodia, Indonesia, Laos, Malaysia, Myanmar, the Philippines, Singapore, Thailand, Timor-Leste, and Vietnam. NWO was defined as the coexistence of a normal BMI (18.5–24.9 kg/m²) and elevated adiposity, determined either by (a) direct body-fat percentage (BF%) measurement using BIA, DXA, or skinfold methods, or (b) validated surrogate indices of central adiposity (e.g., waist circumference, waist–hip ratio, or ≥ 3 metabolic abnormalities per NCEP/ATP III) when BF% data were unavailable. This approach ensured inclusion of regionally representative data where direct body composition tools were not accessible.

We excluded studies focusing solely on special populations (e.g., pregnant women, adolescents, elite athletes, or hospital-based patients with non-generalizable findings). Articles without clearly defined NWO criteria or those that did not report prevalence (or insufficient data to compute it) were also excluded. Only studies published in English and indexed in peer-reviewed journals were considered. Only studies published in English were included because validated translation tools were unavailable for accurate data extraction and quality appraisal at the time of review. Although the search strategy did not impose language restrictions initially, all eligible studies retrieved were English-language publications.

2.3. Information sources and search strategy

We conducted a comprehensive literature search across four major databases: PubMed, EMBASE, Scopus, and Web of Science. To identify grey literature, we also reviewed Google Scholar and manually screened the reference lists of included studies. The search strategy combined Medical Subject Headings (MeSH) and free-text terms related to “normal weight obesity,” “NWO,” “metabolically obese normal weight,” “body fat percentage,” “prevalence,” and “Southeast Asia” (along with individual country names). Boolean operators (AND, OR) were used to enhance search sensitivity and specificity. Detailed search strings and search logs are provided in Appendix A (Supplementary File 1.)

2.4. Study selection process

All records were imported into Rayyan QCRI for screening and deduplication. Two independent reviewers screened titles and abstracts, followed by full-text review of potentially eligible articles. Discrepancies were resolved through discussion or by a third reviewer when needed. If full-text articles were inaccessible, corresponding authors were contacted via email. The study selection process was documented using a PRISMA flow diagram, in line with PRISMA 2020 guidance [16].

A detailed list of studies excluded after full-text review, with corresponding reasons for exclusion, is presented in Appendix B (Table B1).

2.5. Data extraction procedure

A structured Microsoft Excel data extraction form was developed and pilot tested. Two reviewers independently extracted the following variables: author name, year of publication, country and setting, study design, sample size, participant characteristics (age, sex), NWO definition, BMI and body fat cutoffs, measurement method used, and the reported prevalence of NWO (including 95 % confidence intervals when available). If data were disaggregated by sex or age, such subgroups were extracted as well. Extracted data were cross-verified, and discrepancies were resolved by returning to the original articles.

2.6. Quality Assessment and risk of bias

The methodological quality and risk of bias of the included studies were assessed independently by two reviewers using the Joanna Briggs Institute (JBI) Critical Appraisal Checklist for Prevalence Studies [17]. This tool evaluates domains such as sampling frame, recruitment method, measurement validity, statistical analysis, and response rate. Studies were rated as low, moderate, or high risk of bias based on predefined criteria, and disagreements were resolved by consensus.

2.7. Data synthesis and statistical analysis

The primary outcome was the pooled prevalence of NWO in Southeast Asia. A random-effects model (DerSimonian and Laird method) was used to account for expected heterogeneity between studies. Prevalence estimates were stabilized using the Freeman-Tukey double arcsine transformation prior to pooling and back-transformed for interpretability. Statistical heterogeneity was assessed using Cochran's Q test and the I² statistic, with I² >75 % indicating substantial heterogeneity [18,19].

Subgroup analyses were performed based on country, sex, body fat assessment method, and study quality. Where sufficient data (≥10 studies) were available, univariate meta-regression was used to explore the influence of publication year, sample size, and method of fat assessment. Publication bias was assessed visually using funnel plots and statistically using Egger's regression test. All analyses were conducted in R software (version 4.3.1), using the ‘meta’ and ‘metafor’ packages, with sensitivity analyses and graphical outputs generated using STATA version 17.

2.8. Ethical considerations

This systematic review was based on previously published data and did not involve direct interaction with human participants. All included studies stated that ethical clearance was obtained from their respective institutional review boards and were conducted in accordance with international research ethics standards, specifically the Declaration of Helsinki (2013 revision) [20], CIOMS International Ethical Guidelines for Health-related Research Involving Humans (2016) [21], and the Belmont Report (1979) [22].

3. Results

A total of 82 records were identified through a detailed search across five major databases: PubMed [18], EMBASE [16], Scopus [20], Web of Science [15], and Google Scholar [13]. After removing 22 duplicate entries, 60 unique articles were taken forward for title and abstract screening.

During this phase, 34 articles were excluded because they did not match the main objectives of the study—such as lacking data on BF%, focusing on general obesity rather than NWO, or being unrelated to Southeast Asia.

The remaining 26 full-text articles were carefully reviewed to assess whether they met the eligibility criteria. Out of these, 18 articles were excluded for specific reasons: some did not report prevalence data on NWO, others were conducted outside the Southeast Asian region, and a few did not use proper methods to define or measure body fat composition. Additionally, non-original articles like editorials and short communications were also excluded.

Finally, 8 studies that fulfilled all inclusion criteria were considered suitable for the review. These studies provided data relevant to the prevalence of NWO and were included in quantitative synthesis. The detailed selection process is outlined in Fig. 1.

Fig. 1.

PRISMA 2020 flow diagram summarizing the study selection process for systematic review on NWO in Southeast Asia (2010–2025).

3.1. General characteristics of included studies

This systematic review included a total of eight observational studies conducted across five Southeast Asian countries—the Philippines, Malaysia, Thailand, Myanmar, and Singapore—representing the currently available evidence on NWO from the region. These countries were the only ones among the 11 Southeast Asian nations that had published and accessible studies meeting the eligibility criteria during the review period (2010–2025).

The combined study population across all included research comprised 9028 participants, with individual study sample sizes ranging from 40 (Zin et al., Malaysia) to 5920 (Toledano et al., Philippines). However, since NWO is defined only among individuals with normal BMI (18.5–24.9 kg/m²), the effective analytical sample used for meta-analysis included only participants falling within this BMI range. This resulted in a pooled analytic denominator of 5973 normal-BMI individuals, which forms the basis for the prevalence estimates presented in the forest plot.

The studies represented diverse populations such as female university students, community-dwelling adults, wellness center attendees, and working professionals. The mean age of participants ranged from young adults (Foong et al., mean 20.6 years) to middle-aged adults (Annabel Mata et al., mean 53.3 years), while the Yishun Study (Singapore) covered a wide age range (21–90 years). This heterogeneity enabled a broad assessment of NWO prevalence across different demographic strata.

Definitions of NWO were generally aligned with international standards—normal BMI combined with high BF%, typically >30 % for women and >25 % for men (as defined by De Lorenzo et al., 2006 [14] and WHO Expert Consultation, 2004 [15])—while some studies also incorporated metabolic syndrome criteria (e.g., waist circumference or ≥ 3 metabolic abnormalities per NCEP/ATP III), representing population-specific adaptations.

Across studies, the prevalence of NWO ranged from 22.8 % to 82.0 %, influenced by age group, sex distribution, BF% cut-offs, and measurement methods (Tanita, InBody 270, ACCUNIQ BC300, or skinfold). The highest prevalence (82 %) was observed in the Yishun Study (Singapore), where BF%-based classification revealed a substantial proportion of hidden adiposity among individuals with normal BMI, highlighting the limitations of BMI as a standalone screening tool.

All studies were of moderate to high methodological quality, with Quality Assessment (Q.A.) scores between 7/10 and 9/10. Most clearly defined NWO, described sampling methods, and used standardized BF% measurement tools, enhancing the reliability of findings.

This synthesis underscores that while NWO remains under-recognized, it is highly prevalent in Southeast Asia even among individuals with normal BMI, emphasizing the need for region-specific screening protocols and early preventive strategies. A detailed PRISMA-based reporting matrix covering all 27 items for each study is provided in Appendix C (Table C1) to ensure transparency and reproducibility.

Fig. 2 presents the overall pooled prevalence of NWO across eight Southeast Asian studies using a random-effects model. The pooled prevalence was 57 % (95 % CI: 0.37–0.75), indicating that more than half of adults with normal BMI were classified having obesity when BF% or metabolic indicators were applied.

Fig. 2.

Forest plot showing the pooled prevalence of NWO among adult populations in Southeast Asia based on eight studies using a random-effects model.

The analysis demonstrated high between-study heterogeneity (I² = 98.9 %, τ² = 0.0455, p < 0.0001), suggesting that substantial variability exists beyond chance. The τ² value (0.0455) reflects moderate-to-large between-study variance in prevalence estimates, while the I² statistic (98.9 %) confirms that almost all variability arises from true differences among studies rather than sampling error.

Such heterogeneity likely stems from differences in population demographics, measurement tools (BIA, skinfold, surrogate criteria), BF% cut-offs, and study settings. Despite this, the consistently high prevalence across countries underscores that NWO is a widespread but under-recognized condition in Southeast Asia, reinforcing the need for body composition-based screening even among individuals with normal BMI.

Fig. 3 shows the pooled prevalence of NWO stratified by age group across the included Southeast Asian studies. The results reveal a clear age-related gradient in NWO prevalence. Among young adults (18–30 years), the pooled prevalence was 31 % (95 % CI: 0.02–0.73), while among middle-aged adults (31–45 years), prevalence increased sharply to 65 % (95 % CI: 0.44–0.83). The older age group (>45 years), represented by the Yishun Study, showed the highest prevalence at 82 % (95 % CI: 0.78–0.85).

Fig. 3.

Forest Plot Showing Pooled Prevalence of NWO by age group.

The test for subgroup differences was statistically significant (χ² = 33.63, df = 2, p < 0.0001), confirming that age is a major determinant of NWO prevalence. Between-study heterogeneity remained high (I² = 98.9 %, τ² = 0.0455, p < 0.0001), reflecting substantial variability among studies. The relatively lower τ² values within subgroups (τ² = 0.0068 for young, τ² = 0.0096 for middle-aged) suggest that stratification by age reduces heterogeneity to some extent. Thus, the findings indicate that the risk and prevalence of hidden adiposity rise progressively with advancing age, emphasizing the need for early lifestyle interventions and age-specific screening protocols even among individuals maintaining a normal BMI.

Fig. 4 illustrates the pooled prevalence of NWO according to the measurement method used for body fat estimation. Studies using surrogate metabolic criteria (such as waist circumference, waist–hip ratio, or ≥3 metabolic abnormalities) showed a pooled prevalence of 45 % (95 % CI: 0.00–1.00), reflecting broader metabolic definitions of hidden adiposity. In contrast, studies employing bioelectrical impedance analysis (BIA)—the most standardized and direct approach—reported a higher pooled prevalence of 61 % (95 % CI: 0.29–0.88), indicating that direct body composition assessment identifies a larger proportion of individuals with excess fat despite normal BMI. The skinfold-based study (Zin et al., 2014) demonstrated a comparable prevalence of 63 % (95 % CI: 0.39–0.82), consistent with BIA-based results.

Fig. 4.

Forest Plot Showing Pooled Prevalence of NWO by Method used for Measuring NWO.

Heterogeneity remained significant across all subgroups (I² = 98.9 %, τ² = 0.0455, p < 0.0001), but the test for subgroup differences was not statistically significant (χ² = 0.73, df = 2, p = 0.693), suggesting that although measurement method influences prevalence magnitude, it does not fully account for between-study variability. Overall, these findings confirm that high NWO prevalence persists regardless of the assessment technique, underscoring the limitations of BMI alone and the importance of incorporating body composition analysis into clinical and public health screening frameworks.

Fig. 5 displays the pooled prevalence of NWO across five Southeast Asian countries included in the review. The prevalence estimates varied markedly by country, demonstrating substantial regional differences in the burden of hidden adiposity. The highest prevalence was observed in Singapore (82 %; 95 % CI: 0.78–0.85), followed by Myanmar (76 %; 95 % CI: 0.69–0.81), indicating a high proportion of individuals with normal BMI but elevated BF%. Malaysia showed a pooled prevalence of 52 % (95 % CI: 0.10–0.86), while Thailand reported 47 % (95 % CI: 0.40–0.53). The Philippines exhibited the lowest combined prevalence at 45 % (95 % CI: 0.00–1.00).

Fig. 5.

Forest plot showing pooled prevalence of NWO by country.

Between-country heterogeneity was statistically significant (χ² = 102.93, df = 4, p < 0.0001), confirming that national differences significantly influenced prevalence estimates. Overall heterogeneity remained high (I² = 98.9 %, τ² = 0.0455, p < 0.0001), reflecting variability in study populations, measurement methods (BIA, skinfold, surrogate criteria), and BF% thresholds.

Collectively, these findings suggest that NWO is highly prevalent throughout Southeast Asia, though the magnitude varies by country, likely reflecting diverse lifestyle patterns, nutritional transitions, and population-level adiposity profiles. The consistently high prevalence across nations underscores the urgent need for country-specific screening and preventive strategies targeting hidden obesity in normal-BMI adults.

Fig. 6 presents the pooled prevalence of NWO stratified by sex across the included studies. The pooled prevalence among males was 63 % (95 % CI: 0.24–0.94), while among females, it was 51 % (95 % CI: 0.17–0.84). Although the prevalence was slightly higher in men, the test for subgroup differences was not statistically significant (χ² = 0.51, df = 1, p = 0.4731), indicating that NWO affects both sexes to a similar extent in Southeast Asian populations.

Fig. 6.

Forest Plot Showing Pooled Prevalence of NWO by Sex (Male vs Female).

High between-study heterogeneity was observed in both subgroups (I² = 99.3 % for males; I² = 98.2 % for females; τ² = 0.0455, p < 0.0001), suggesting that study-level differences—such as sample characteristics, measurement devices, and BF% cut-off criteria—contributed to variability in prevalence estimates.

Overall, these findings highlight that NWO is a prevalent condition in both men and women with normal BMI, reinforcing the need for sex-inclusive screening programs using body composition measurements rather than BMI alone. The absence of a significant sex difference suggests that metabolic and adiposity risks associated with NWO are widespread across both sex in the Southeast Asian region.

Fig. 7 shows the results of the sensitivity analysis limited to studies that defined NWO strictly using direct BF% measurements. The pooled prevalence among these BF%-based studies was 61 % (95 % CI: 0.36–0.83), indicating that nearly two-thirds of adults with normal BMI were classified having obesity when BF% criteria were applied.

Fig. 7.

Forest plot showing pooled prevalence of NWO based on BF%-Defined studies (sensitivity analysis).

Among individual studies, the prevalence ranged from 23 % (Foong et al., 2010, Malaysia) to 82 % (Yishun Study, 2021, Singapore), demonstrating wide variability likely due to demographic and methodological differences. Despite the narrower inclusion criteria, between-study heterogeneity remained high (I² = 98.6 %, τ² = 0.0698, p < 0.0001), suggesting residual variation in study design, BF% cutoff thresholds, and measurement devices (Tanita, InBody, ACCUNIQ).

The similarity of this pooled estimate to that of the main model (57 %) indicates that the overall results are robust and not driven by studies using surrogate definitions of NWO. This confirms that high hidden adiposity prevalence in Southeast Asian adults persists even when analyses are confined to studies employing direct BF%-based definitions, reinforcing the reliability of the main findings.

Fig. 8 depicts the funnel plot assessing potential publication bias among the six studies that defined NWO using BF% criteria. The plot shows a relatively symmetrical distribution of studies around the pooled estimate on the Freeman–Tukey transformed scale, indicating no major evidence of publication bias or small-study effects.

Fig. 8.

Funnel plot for publication bias in sensitivity analysis (BF%-Based studies only).

Although minor asymmetry is observed, it is likely attributable to the limited number of included studies and high between-study heterogeneity (I² = 98.6 %). Given the small sample of studies (<10), formal statistical tests such as Egger's regression were not applied, as they would lack power to detect bias reliably.

This supports the robustness of the pooled prevalence estimate obtained in the sensitivity analysis.

Because fewer than ten studies defined NWO using body-fat percentage, formal statistical tests for publication bias (e.g., Egger's regression) were not applied, in accordance with established methodological guidance that such tests lack power and yield unreliable results with small study numbers.

Fig. 9 illustrates the domain-wise risk of bias assessment for the eight studies included in this review. The evaluation was based on five methodological domains, including the sampling process (D1), deviations from intended intervention (D2), missing outcome data (D3), measurement of outcomes (D4), and selective reporting (D5). Most studies demonstrated low risk across multiple domains, particularly in D1 and D2, indicating clearly described sampling methods and adherence to intended measurement protocols.

Fig. 9.

Domain-wise risk of bias assessment of included studies on NOW.

However, several studies exhibited potential concerns. For example, Johari et al. (2017) lacked sufficient information on the sampling process and showed concerns related to missing data. Kobayashi et al. (2023) and Thant Zin et al. (2021) showed high risk in outcome measurement and reporting, which may influence the accuracy of body fat assessment or classification of NWO. Toledano et al. (2022), although nationally representative, was marked high risk in both sampling and reporting domains. Zin et al. (2014) also had limitations in representativeness and transparency.

In contrast, Annabel Mata et al. (2017) and the Yishun Study (2021) showed relatively strong methodological quality, with low risk across most domains. Overall, while the evidence base is generally robust, some studies carry methodological limitations that should be considered when interpreting the pooled estimates of NWO prevalence in the Southeast Asian context.

Subgroup analyses (Fig. 10, Fig. 11) were performed based on the classification system adopted by individual studies as outlined in Table 3 (see Table 2). Fig. 10 presents the pooled prevalence of normal weight obesity (NWO) among studies that applied the Asian BMI definition (18.5–22.9 kg/m²). Using a random-effects model (REML, Freeman–Tukey transformation), the pooled prevalence was 46 % [95 % CI 0.10–0.84], indicating that nearly half of individuals with a normal BMI by Asian cut-offs had excess body fat consistent with NWO.

Fig. 10.

Forest plot of pooled prevalence of normal weight obesity using Asian BMI definition (18.5–22.9 kg/m²).

Fig. 11.

Forest plot of pooled prevalence of normal weight obesity using western BMI definition (18.5–24.9 kg/m²).

Table 3.

Comparison between Asian-specific and international (WHO) BMI classification criteria.

BMI Category	WHO International Classification (kg/m2)	WHO Asian-Specific Classification (kg/m2)	Interpretation/Risk Description
Underweight	<18.5	<18.5	Risk of nutritional deficiency and low body fat
Normal weight	18.5–24.9	18.5–22.9	Healthy weight range for most individuals
Overweight (At Risk)	–	23.0–24.9	Elevated metabolic risk despite normal BMI by Western cut-offs
Obese I	25.0–29.9	25.0–29.9	Increased risk of metabolic complications
Obese II	30.0–34.9	≥ 30.0	High risk of comorbidities
Obese III (Morbid obesity)	≥35.0	≥35.0	Very high risk of metabolic and cardiovascular disease

Source: WHO Expert Consultation. Appropriate Body Mass Index for Asian Populations and Its Implications for Policy and Intervention Strategies. Lancet 2004; 363: 157–163 [15].

Table 2.

Summary of included studies reporting the prevalence of NWO among adults in Southeast Asia.

Author (Year)	Country	Sample Size (Total)	Mean Age/Age Range	Study Population	BF% Measurement Method	NWO Cutoff Criteria/Definition	Prevalence of NWO	Quality Score from 10 Points (Q.A.)
Annabel Mata et al. (2017) [23]	Philippines	1367 (Analytic n = 630)	Mean 53.3 (18–86)	Adults attending wellness center, Manila	Surrogate (Metabolic Criteria)	BMI 18.5–22.9 + ≥3 metabolic abnormalities (NCEP/ATP III)	28.4 % (179/630)	08
Foong et al. (2010) [24]	Malaysia	381 (Analytic n = 294)	Mean 20.6 ± 1.6	Female university students	Bioelectrical Impedance Analysis (BIA, Tanita)	BMI 18.5–22.9 + BF% ≥30 % (Female)	22.8 % (67/294)	07
Johari et al. (2017) [25]	Malaysia	327	Mean 50.4 ± 5.6	Urban women attending screening camps	Bioelectrical Impedance Analysis (InBody 270)	Normal BMI + BF% >33 % (Female)	72.8 % (238/327)	08
Kobayashi et al. (2023) [26]	Thailand	250	Mean 20.2 ± 1.1	Thai female university students	BIA (ACCUNIQ BC300)	BMI 18.5–24.9 + BF% >30 % (Female)	46.8 % (117/250)	09
Toledano et al. (2022) [27]	Philippines	5920 (Analytic n = 3710)	Mean 43.5 ± 15.2 (≥20 yrs)	Adults from national nutrition survey	Surrogate (WC/WHR Criteria)	BMI 18.5–24.9 + WC ≥ 90 cm (M)/≥80 cm (F) or WHR ≥0.90/0.85	62.7 % (2325/3710)	09
Zin et al. (2014) [28]	Malaysia	40 (Analytic n = 19)	Mean 20.2 ± 1.3	Year 2 medical students	Skinfold (Tanita conversion equations)	BMI 18.5–22.9 + BF% ≥20.35–24.13 % (Study-defined percentile)	63.0 % (12/19)	07
Thant Zin et al. (2021) [28]	Myanmar	201	Mean 23.5	University staff with normal BMI	BIA (Tanita)	BMI 18.5–24.9 + BF% >25 % (M)/>35 % (F)	75.6 % (152/201)	08
Yishun Study (2021) Chen at al [29]	Singapore	542	Age 21–90 yrs	Community-dwelling adults, Yishun	BIA (InBody 770)	BF% >25 % (M)/>35 % (F)	82.0 % (444/542)	09

Between-study variability was high (I² = 98.7 %, τ² = 0.0652, p < 0.001), reflecting substantial heterogeneity across included studies. Individual study estimates ranged from 22.8 % (Foong 2010) to 72.8 % (Johari 2017), likely due to differences in population age, adiposity measurement methods (BIA vs skinfold vs metabolic surrogates), and sampling settings.

Despite heterogeneity, the overall pooled estimate suggests that a large proportion of Asian adults with BMI <23 kg/m² still exhibit metabolic obesity, reinforcing that BMI alone underestimates true adiposity risk in Asian populations.

Fig. 11 shows the pooled prevalence of normal weight obesity (NWO) among studies that used the Western BMI definition (18.5–24.9 kg/m²). The random-effects model (REML, Freeman–Tukey transformation) estimated a pooled prevalence of 67 % [95 % CI 0.41–0.89], indicating that approximately two-thirds of individuals classified as normal-weight within this BMI range exhibited elevated body fat consistent with NWO. Between-study variability remained high (I² = 97.8 %, τ² = 0.0286, p < 0.001), suggesting considerable heterogeneity in effect sizes across included studies. The prevalence values across studies ranged from 47 % (Kobayashi 2023) to 82 % (Chen 2021), with larger sample studies such as Toledano et al. (2022) exerting greater influence on the pooled estimate. The diamond representing the summary estimate is positioned toward the upper range of the plot, visually confirming a higher pooled NWO prevalence among populations assessed with the Western BMI definition. The narrow confidence intervals of larger studies and overlapping CIs across most studies indicate a consistent overall pattern despite the observed heterogeneity.

Fig. 12 displays the influence diagnostics for each study included in the meta-analysis, assessing their individual effect on the pooled prevalence of normal weight obesity (NWO) and on overall model heterogeneity. The plots illustrate standardized residuals, Cook's distance, covariance ratio, and changes in τ² and Q statistics when each study is omitted. Across indices, most studies cluster within acceptable limits, indicating model stability and the absence of severe outliers. However, a few studies—particularly Mata et al. (2017), Toledano et al. (2022), and Chen et al. (2021)—show relatively higher influence values on τ² and Cook's distance, suggesting that they contribute more strongly to the observed heterogeneity. Despite these localized effects, no study demonstrated disproportionate leverage or distortion of the pooled estimate, confirming that the overall meta-analytic findings remain robust after accounting for influential data points.

Fig. 12.

Influence diagnostics plot for individual studies in the NWO meta-analysis.

Fig. 13 presents the enhanced Baujat plot illustrating each study's contribution to overall heterogeneity (x-axis) and its influence on the pooled prevalence of normal weight obesity (y-axis). Most studies cluster toward the lower-left quadrant, indicating limited impact on between-study variance and pooled effect size. In contrast, Mata et al. (2017), Toledano et al. (2022), and Chen et al. (2021) appear in the upper-right region of the plot, signifying that these studies exert the greatest influence on both the heterogeneity (Q-statistic) and overall pooled estimate. Their higher sample sizes and differing methodological approaches likely explain their prominent positions. Overall, the plot confirms that while a few studies contribute disproportionately to heterogeneity, the influence pattern is not dominated by a single outlier, supporting the stability and generalizability of the pooled meta-analytic results.

Fig. 13.

Baujat plot showing study contributions to heterogeneity and influence on overall effect size.

Table 4 presents the meta-regression findings exploring study-level factors that accounted for between-study heterogeneity in NWO prevalence.

Table 4.

Meta-regression of study-level moderators explaining heterogeneity in NWO prevalence.

Moderator	β (Estimate)	SE	z-value	p-value	R2 analog (%)	Interpretation
BMI definition (Asian vs Western)	−0.18	0.07	−2.53	0.02 ∗	22	Lower pooled prevalence when Asian BMI cut-off (18.5–22.9 kg/m2) used
Measurement method (BIA vs others)	0.15	0.06	2.36	0.03 ∗	17	Higher prevalence with BIA-based body-fat assessment
Mean age (years)	0.011	0.005	2.1	0.04 ∗	10	Prevalence increases modestly with age
Population type	–	–	NS	0.29	3	Not significant
Sample size (log n)	–	–	NS	0.41	2	Not significant
Publication year	–	–	NS	0.48	1	Not significant
Overall Model (BMI def + method + mean age)	–	–	–	–	47 %	Combined moderators explained nearly half of between-study heterogeneity

The multivariate model including BMI definition, measurement method, and mean age substantially reduced between-study variance (τ² = 0.033 vs 0.063 in the null model), explaining about 47 % of total heterogeneity (R² analog = 47 %).

Among the moderators, BMI definition was a significant predictor (p = 0.02); studies adopting the Asian cut-off (18.5–22.9 kg/m²) reported lower pooled prevalence compared with those using the Western cut-off (18.5–24.9 kg/m²).

Measurement method also showed a significant positive association (p = 0.03), indicating that studies employing bioelectrical-impedance analysis (BIA) produced higher prevalence estimates than those using skinfold or surrogate metabolic criteria.

Mean participant age had a modest but significant effect (p = 0.04), with older populations demonstrating slightly greater NWO prevalence.

Other variables—population type, sample size, and publication year—were not statistically significant (all p > 0.05; R² ≤ 3 %), suggesting minimal contribution to heterogeneity.

Overall, these results indicate that methodological and demographic differences across studies account for nearly half of the observed heterogeneity, while the direction of effects remains consistent across the included evidence base.

4. Discussion

This systematic review and meta-analysis identified a pooled prevalence of 57 % (95 % CI: 37 %–75 %) for normal weight obesity (NWO) among adults in Southeast Asia. This phenotype—where individuals possess a normal BMI but elevated body fat percentage BF%—has substantial clinical and public health implications, particularly in populations predisposed to central adiposity and insulin resistance despite appearing lean. The prevalence of NWO varied widely across studies (22.8 %–82 %), likely reflecting differences in BF% cut-offs, assessment methods (Tanita, InBody, DXA), age structures, and sex distribution. However, even the lowest estimates far exceed those reported in Caucasian populations, where prevalence typically ranges between 15 % and 30 %. This supports the well-documented ethnic differences in body composition, where Asian populations tend to accumulate visceral fat at lower BMI levels, increasing their susceptibility to metabolic complications.

In Vietnam, Binh et al. [30] reported that 45 % of normal-weight adults exhibited elevated BF% and metabolic abnormalities, mirroring findings from Myanmar and the Singapore Yishun Study [29], where BMI missed over 70 % of high-risk individuals. Similar trends have been documented in Malaysia and Thailand, particularly among university women [24,31], where behavioral, dietary, and lifestyle changes contribute to disproportionate adiposity despite normal total body weight. In the Philippines [23], national-level data identified that 62.7 % of adults with normal BMI met criteria for central obesity using WC and WHR, corroborating the invisibility of metabolic risk under BMI classification. Comparative evidence from East Asia further supports this trend. In China, Xu et al. (2021) [32] reported a 3–5-fold increased risk of diabetes among adults with normal BMI but high body fat percentage, while Kim et al. (2014) [33] in Korea and Japanese cohort [34] studies have also demonstrated that metabolic syndrome and cardiovascular risk can occur in individuals who appear lean by BMI criteria. These findings suggest that the discordance between BMI and true adiposity is not limited to Southeast Asia but represents a broader pan-Asian phenomenon where ethnic-specific fat distribution patterns elevate risk even within the so-called normal BMI range.

The conceptual limitations of BMI as a surrogate for adiposity are clearly evident in this review. BMI, being a weight-to-height ratio, does not differentiate between fat and lean mass, nor does it reflect visceral fat accumulation—the critical determinant of metabolic risk in Asian populations. As a result, reliance on BMI alone can lead to both under-diagnosis and over-diagnosis: under-diagnosis among metabolically obese individuals with normal BMI, and over-diagnosis among muscular individuals with high BMI but low-fat mass. The WHO Expert Consultation (2004) [15] recognized these shortcomings, recommending lower BMI thresholds (23.0 and 27.5 kg/m²) for defining overweight and obesity in Asian populations. However, our findings demonstrate that even with these Asian-specific BMI cut-offs, a significant proportion of individuals remain undiagnosed for hidden adiposity. This reinforces the WHO's own conclusion that BMI is an inadequate standalone measure for assessing metabolic health in Asian populations and should be complemented by direct or surrogate measures of body composition.

The discordance between BMI and true adiposity has profound diagnostic consequences. Individuals classified as “normal weight” by BMI may already harbor high visceral fat and metabolic abnormalities, thus representing a high-risk but undiagnosed subgroup. Conversely, individuals with higher muscle mass—such as athletes or manual laborers—may be erroneously categorized as overweight or obese by BMI, leading to potential over-diagnosis. Such misclassification undermines the sensitivity and specificity of BMI-based screening tools and may misguide both clinical decision-making and public health surveillance.

Parallel evidence from other Asian nations reinforces the consistency of these findings. In Japan, the prevalence of NWO has been reported between 30–45 %, and individuals within the normal BMI range exhibited elevated insulin resistance and triglyceride levels [34]. In South Korea, Kim et al. (2014) [33] documented a high proportion of normal-weight adults with metabolic syndrome components, while in China, Xu et al. (2021) [32] identified significantly higher diabetes and dyslipidemia risk in NWO individuals compared to truly lean counterparts. Collectively, these findings affirm that BMI-based classification systems underestimate true metabolic risk across Asia, regardless of subregion, diet, or socioeconomic context.

From a policy perspective, these findings signal the need for restructuring screening frameworks across Asia to move beyond BMI-centric surveillance. Despite WHO's 2004 [15] recommendation for Asian-specific BMI cut-offs, our pooled prevalence data show that metabolic risk remains underestimated. Incorporating body composition metrics—such as bioelectrical impedance (BIA), waist circumference, or waist–hip ratio—into national screening programs would provide a more accurate reflection of risk. Countries like Singapore (“Screen for Life”) [29] and Malaysia (MyHEALTH program) [35] have begun integrating waist circumference into health assessments, but region-wide adoption of such measures remains limited. Routine use of portable BIA devices in community health programs could enhance early detection of hidden adiposity, particularly in young adults and women—groups found to have higher undetected risk. Policymakers should consider defining region-specific BF% thresholds (≥25 % for men, ≥30 % for women) and standardizing measurement methods to ensure comparability across surveillance systems. Ultimately, reorienting obesity surveillance and intervention strategies toward body composition-based screening could help identify high-risk individuals earlier, enabling preventive lifestyle and dietary interventions before overt metabolic disease develops.

Although a funnel plot was generated for BF%-defined studies, Egger's regression test was not applied because fewer than ten studies met the inclusion criteria for quantitative synthesis. As recommended by Cochrane and Sterne et al. (2011) [36,37], regression-based bias tests lack statistical power with small samples and may yield misleading results. The decision not to perform Egger's test was therefore methodological and consistent with best-practice guidelines.

This meta-analysis provides robust evidence that BMI alone substantially underestimates the burden of obesity and metabolic risk in Asian populations. The consistently high prevalence of NWO across Southeast and East Asia highlights the urgent need to revise screening frameworks, emphasizing composition-based rather than weight-based assessments. Adopting such an approach could prevent missed diagnoses, improve early intervention, and more accurately reflect the region's cardiometabolic disease burden.

4.1. Strengths

This systematic review is among the first to provide a pooled prevalence estimate of NWO in Southeast Asia using a rigorous methodology. It adheres to PRISMA 2020 guidelines, with transparent reporting of eligibility criteria, database coverage, and a structured risk of bias assessment. The inclusion of eight observational studies across five countries with a total sample size of 9028 participants enhances the regional relevance and statistical robustness of the analysis.

A notable strength is the focus on BF% as the diagnostic criterion for obesity, rather than relying solely on BMI. This addresses a critical methodological gap in previous studies and reflects the more accurate detection of cardiometabolic risk in Asian populations. The synthesis also incorporated subgroup comparisons by country and sex and triangulated findings with both regional and international literature, improving contextual interpretation.

Furthermore, the meta-analytic approach used the Freeman-Tukey double arcsine transformation to stabilize variance in proportion data and accounted for between-study heterogeneity using a random-effects model. Risk of publication bias was formally assessed via funnel plots and Egger's test, adding to the methodological credibility of the findings.

4.2. Limitations

Despite its strengths, the review has several limitations. First, the substantial heterogeneity (I² = 98.9 %) limits the generalizability of the pooled estimate. This heterogeneity likely stems from variations in NWO definitions, body fat assessment tools (e.g., BIA vs. DXA vs. skinfold), and sample characteristics (e.g., age, sex, and urban-rural composition).

Second, only five of the eleven Southeast Asian countries were represented (Philippines, Malaysia, Singapore, Thailand, Myanmar), leaving significant geographical gaps. Countries such as Indonesia, Cambodia, Laos, Brunei, Timor-Leste, and Vietnam either had no eligible studies or lacked full-text availability, which may skew regional estimates and underrepresent certain at-risk populations.

Third, the cross-sectional nature of the included studies limits causal inferences between NWO and long-term health outcomes. While the review identifies high NWO prevalence, it cannot assess incident diabetes, cardiovascular events, or mortality among this group. Longitudinal studies are needed to establish such associations.

Additionally, most studies focused on adults aged 18–60, with minimal inclusion of elderly populations, who may exhibit different body fat distribution and metabolic profiles. Finally, while quality appraisal was conducted using the JBI checklist, some studies had incomplete reporting on sampling methods, measurement protocols, or response rates, introducing potential bias in outcome classification.

Differences in body fat assessment methods (DXA vs BIA), device calibration, and hydration effects may have introduced measurement variability, and therefore, cross-study comparisons should be interpreted with caution.

5. Conclusion

The present review highlighted a consistently high prevalence of NWO among adults across Southeast Asia, demonstrating that a substantial proportion of individuals classified as normal weight by BMI criteria may still exhibit adverse body composition profiles. This concealed adiposity poses a significant risk for cardiometabolic complications, even in the absence of overt obesity. The findings strongly suggest that BMI alone is inadequate for accurately identifying individuals at metabolic risk within these populations.

The review also underscored the increased vulnerability of specific subgroups—particularly women and younger adults—to undetected NWO. The use of direct adiposity measures such as BF% was shown to be more reflective of metabolic risk, supporting the inclusion of such metrics in routine clinical and public health screening frameworks. Moreover, the wide variability in prevalence estimates across studies reflects the methodological inconsistencies and the urgent need for region-specific, standardized diagnostic criteria for NWO.

There is a clear implication for health systems to transition from reliance on BMI-centric models toward a more nuanced, composition-based assessment approach. Integrating such practices into preventive health strategies may facilitate earlier detection and intervention, ultimately mitigating the long-term burden of non-communicable diseases associated with hidden adiposity.

CRediT authorship contribution statement

K.G. Sruthi: Writing – review & editing, Writing – original draft, Data curation, Conceptualization. C. Aditya: Validation. Paramjot Panda: Writing – original draft, Visualization, Software, Methodology, Formal analysis. Jyoti Ranjan Mohanty: Writing – review & editing, Writing – original draft. Manas Ranjan Behera: Writing – review & editing, Writing – original draft, Validation.

Appendix A. Supplementary data

The following are the Supplementary data to this article: