Introduction

Obesity has become a major global health concern, traditionally assessed and classified using the body mass index (BMI) (1,2). While BMI provides a quick measure of excess weight, it does not account for the relative proportions of fat and muscle mass, nor does it capture fat distribution (3). As a result, individuals with the same BMI may have vastly different body composition profiles, leading to potential misclassification of their metabolic and cardiovascular risk (3-5). Considering these limitations, the focus on body composition assessment has grown, placing emphasis on more detailed measurements. Among the available tools, bioelectrical impedance analysis (BIA) has emerged as a practical, non- invasive, and cost-effective method to evaluate fat mass, muscle mass, and distribution patterns, offering a more nuanced perspective on overall health (6,7).

Beyond distinguishing between various tissue components, body composition metrics such as phase angle (PhA) have gained attention for their ability to shed light on cellular health and integrity (7,8). PhA, derived from the reactance and resistance measurements in BIA, reflects the overall function and viability of cell membranes. Higher PhA values often correlate with healthier cell membranes and better nutritional status, whereas lower values can indicate compromised cell integrity and heightened morbidity risk (8-10). By providing an additional layer of information about the body’s physiological state, PhA complements traditional body composition parameters, expanding the insights available for clinicians and researchers in understanding obesity-related risks (11,12).

Given the growing necessity for a more comprehensive and accurate obesity assessment, multiple classification systems have been proposed, notably those from the World Health Organization (WHO) and the National Health and Nutrition Examination Survey (NHANES) (13). However, inconsistencies in how these systems categorize weight status underscore the need for harmonized or comparative approaches (4,14-16). Therefore, the objective of this study is to critically evaluate the discrepancies between the WHO and NHANES classification systems in terms of how they define obesity in a Colombian cohort. We will also analyze the relationship between phase angle (PhA), BMI and other body composition variables, aiming to contribute to a more robust and precise framework for assessing obesity and body composition within this specific context.

Methods

Results

A total of 3,255 individuals were included, of whom 74.3% (n = 2,419) were women. The median age for the overall sample was 42 years(IQR 33-53), with a median BMI of 28.0 kg/ m² (IQR 24.7-32.0). Key body composition parameters included a median fat mass index (FMI) of 10.9 kg/m² (IQR 8.2-14.35) and a median phase angle (PhA) of 5.4° (IQR 5.0-5.9). Table 1 provides a comprehensive summary of these and other body composition parameters. When stratified by sex, significant differences emerged in weight (p < 0.001, BMI (p < 0.001), body fat mass (p < 0.005), and visceral adipose tissue (p < 0.001)), whereas fat-free mass, skeletal muscle mass, PhA, and waist-to-hip ratio showed no statistically significant variation (p > 0.05). Overall, although women tended to have lower median weight values than men, they displayed comparable FFM, SMM, and PhA levels. A comprehensive overview of these findings is presented in Table 2 and Figure 1.

Figure 1. : Comparison of BMI, FMI, PhA, and SMI by sex. Note. Box plots comparing body mass index (BMI), fat mass index (FMI), phase angle (PhA), and skeletal muscle index (SMI) between women (green) and men (blue). Statistical significance was assessed using the Mann-Whitney U test. Significant differences were observed for BMI (p = 0.000) and FMI (p = 0.000), with men showing higher median values for both. No significant differences were observed for PhA (p = 0.956) or SMI (p= 0.178), indicating similar distributions between sexes for these variables. These results highlight sex-base differences in adiposity measures but not in muscle-related or functional indicators.

Source: The authors.

Chi-squared analyses revealed a significant global association between the WHO and NHANES obesity classifications (χ² = 8100.88, p < 0.001). The agreement between these two classification systems was quantified using Cohen’s kappa coefficient (κ = 0.39), indicating a fair level of concordance across the entire study cohort. However, a closer look at the individual categories revealed notable differences in how patients were classified under each system. For instance, individuals classified as “Underweight” by WHO were more frequently deemed “Normal” (34.2%) or “Mild Fat Deficit” (32.9%) by NHANES, while those labeled “Normal” under WHO mostly retained the same classification (72.8%) but had a sizable proportion (22.7%) flagged as “Excess Fat.” Among individuals classified as “Overweight” by the WHO system, the majority (73.1%) shifted to “Excess Fat” under NHANES, while 18.9% were categorized as “Obesity Class I.” A similar pattern was observed in higher obesity classes. WHO “Obesity Class I” correlated strongly with NHANES “Obesity Class I” (72.3%) and, to a lesser extent, “Obesity Class II” (18.0%). Likewise, WHO “Obesity Class II” largely overlapped with NHANES “Obesity Class II” (75.7%) and “Obesity Class III” (18.7%), whereas WHO “Obesity Class III” was mirrored by NHANES “Obesity Class III” in 91.7% of cases. A more comprehensive breakdown of these discrepancies is shown in Table 3, while Figure 2 offers a visual overview. Interestingly, when stratified by sex, the concordance was higher in females (κ = 0.43, moderate) compared to males (κ = 0.28, fair), which contrasts with the fair global concordance observed in the overall cohort.

Figure 2. : Proportion of NHANES classifications by WHO classification Note. Groupedbar chart displaying the proportion of NHANES body fat classifications within each WHO BMI classification category. Each bar represents the distribution of NHANES categories (Severe Fat Deficit, Moderate Fat Deficit, Mild Fat Deficit, Normal, Excess Fat, Obesity Class I, II, and III) within a specific WHO classification (Underweight, Normal, Overweight, Obesity Class I, II, and III). The chart highlights notable discrepancies between the two systems, particularly in the extremes, where classifications diverge significantly. These results underscore the limitations of BMI-based categorizations in capturing nuanced differences in body composition.

Global correlation analysis revealed the strongest associations (ρ > 0.70, p < 0.001) between BMI and body fat mass (BFM) (global ρ = 0.929), and between BFM and visceral adipose tissue (VAT) (global ρ = 0.978). These patterns held true across sexes, with women displaying correlations of ρ = 0.955 and ρ = 0.921, respectively, and men showing similar strengths (ρ = 0.939 and ρ = 0.914, respectively), consistently classified as very strong. In contrast, correlations involving BMI and FFM, SMM, waist-to-hip ratio (WHR), or phase angle (PhA) were negligible (ρ < 0.30) or nonsignificant (p > 0.05) in all analyses. Moderate correlations (0.50 µ ρ < 0.70, p < 0.001) were observed between FFM or SMM and PhA, as well as with WHR in the overall sample. Notably, the correlation between FFM and WHR was classified as moderate globally (ρ = 0.502) but low in men (ρ = 0.493), reflecting subtle sex-specific differences. Collectively, these findings underscore that the most robust correlations are observed among measures of adiposity (BMI, BFM, VAT) and closely related parameters, while associations with WHR and PhA remain weaker, highlighting the multifaceted nature of body composition. Further details of these analyses are provided in Table 4.

In the multivariable linear regression model using fat mass index (FMI) as the dependent variable, overall explanatory power was modest (R² = 0.950, F = 15291.27, p < 0.0001, RMSE= 1.0972). Phase angle (PhA) did not reach statistical significance (β = -0.0208, p = 0.444), whereas BMI (β = 0.7508, p < 0.0001), age (β = 0.0113, 95% CI 0.008-0.014, p < 0.0001) and sex (β = -3.3379, 95% CI -3.425 to -3.251, p < 0.0001) emerged as strong predictors of FMI.

In contrast, the model for skeletal muscle index (SMI) demonstrated a considerably higher explanatory capacity (R² = 0.332, F = 403.7, p < 0.0001, RMSE = 1.0379). Here, PhA was highly significant (β = 1.032, 95% CI 0.982-1.083, p < 0.0001), whereas neither BMI (p = 0.586), age (p = 0.812) nor sex (p = 0.095) contributed meaningfully to the SMI variance. Taken together, these findings suggest that while BMI, age, and sex play pivotal roles in predicting adiposity (FMI), PhA exerts a negligible effect on fat mass but is a strong determinant of muscle mass (SMI). A detailed representation of these findings is provided in Figure 3.

Figure 3.: Partial effects of phase angle, age, and sex on fat mass index (FMI) and skeletal muscle index (SMI) Note. Scatterplots illustrating the partial effects of phase angle (PhA), BMI, age, and sex on fat mass index (FMI) (top row) and skeletal muscle index (SMI) (bottom row) in the global analysis. The dashed lines represent the fitted trends for each predictor. The first column shows the effect of PhA, highlighting its significant influence on SMI but not FMI. The second column depicts the effect of BMI, illustrating its strong association with FMI. The third column represents the effect of age, and the fourth column captures the effect of sex, where 0 = Female and 1 = Male. These visualizations demonstrate the distinct relationships of these variables with fat and muscle compartments, supporting the importance of advanced body composition analysis in clinical assessments.

Source: The authors.

Discussion

This comprehensive analysis provides valuable insights into the relationship between traditional anthropometric measurements and advanced body composition parameters. Notably, while women exhibited lower median weight values compared to men, they demonstrated comparable levels of FFM, SMM, and PhA. These findings could challenge traditional assumptions about sex- based differences in body composition and may be particularly relevant in the Latin American context, specifically Colombia, where cultural factors such as increased physical activity levels and dietary patterns might contribute to these outcomes (13,18-20). Further research is warranted to better understand these observations, particularly in the context of cultural and regional factors that may influence body composition patterns in Latin America. Clinically, these results underscore the need for personalized treatment approaches that account for sex-specific differences in order to improve obesity management. Similar findings have been reported in previous studies, which suggest that physiological differences may significantly influence metabolic health and responses to obesity treatments (21).

The analysis of classification system concordance between WHO and NHANES revealed intriguing patterns with important clinical implications. While the overall agreement was fair, the higher concordance observed in females compared to males suggests that current classification systems may be better calibrated for women’s body composition patterns (18). However, this observation could partly be explained by the larger proportion of women in our sample, which could have influenced the results. Notably, the discrepancies observed, particularly at the extremes of the distribution, underscore the limitations of BMI-based categorization systems (3,22). These findings align with growing evidence that BMI, while useful as a population-level screening tool, may not adequately capture individual variations in body composition, especially in populations with unique anthropometric characteristics (2,3,5).

The correlation analyses provided robust evidence for the relationship between different body composition parameters. The strong correlations observed between BMI and BFM (ρ = 0.929) validate the utility of BMI as a surrogate marker for adiposity in this population. In contrast, the moderate to weak correlations between BMI and measures of lean mass (FFM, SMM) reinforce the inherent limitations of BMI in assessing muscular components and structural body composition. These findings are particularly relevant in the clinical setting, where accurate assessment of both fat and lean mass is essential for tailoring interventions and monitoring treatment outcomes (23-25). For instance, clinicians could leverage advanced technologies that offer more comprehensive data on muscle mass and quality, facilitating the development of more precise, personalized, and effective treatment strategies.

Perhaps the most striking findings emerged from the multivariable regression analyses. The robust explanatory power of the FMI model (R² = 0.950) compared to the moderate predictive capacity of the SMI model (R² = 0.332), reveals fundamental differences in how body composition parameters interact. Specifically, BMI proved to be a strong predictor of fat mass, whereas phase angle (PhA) showed a pronounced association with muscle mass (β = 1.032, p < 0.0001) and minimal relevance for adiposity. Interestingly, BMI did not meaningfully correlate with muscle mass, underscoring the limitations of relying solely on BMI to evaluate muscular health.

Furthermore, age and sex emerged as significant contributors to FMI but did not meaningfully influence SMI. Taken together, these findings underscore the importance of considering BMI, age, and sex when evaluating fat mass, while highlighting PhA as a valuable tool for assessing muscle mass-a component often overlooked in traditional obesity assessments (23,26,27). Supporting the development of age- specific and sex-specific guidelines for diet and physical activity becomes essential under this paradigm. Clinically, this means that lifestyle recommendations should be tailored not just based on overall health status but also on these demographic characteristics, which could impact the efficacy of diet and exercise regimens (28).

The noted discrepancies between the WHO and NHANES systems, along with the varying correlations among different body composition metrics call for a reevaluation of current obesity classification standards. By emphasizing the significance of phase angle as a determinant of muscle mass, we underscore the necessity of incorporating functional markers in the evaluation of obesity-beyond traditional indices like BMI. Our findings highlight the need for more inclusive criteria that reflect diverse body composition profiles, especially those in underrepresented populations like Colombians. Clinically, this reevaluation could lead to more accurate risk assessment and personalized treatment pathways. By incorporating both WHO and NHANES criteria along with emerging biomarkers into a multifaceted classification approach, healthcare providers can better identify individuals at risk and tailor treatments to specific conditions like insulin resistance or hypertension (20,29,30). This is especially relevant in populations where standard anthropometric assumptions may not apply, as demonstrated by the similar muscle mass profiles observed between men and women in our study.

For example, patients with low muscle mass at high risk of insulin resistance may benefit from a dual-intervention approach: dietary modifications aimed at lowering glycemic loads and targeted pharmacotherapy to address metabolic dysfunction, complemented by progressive resistance training protocols designed to increase skeletal muscle mass, thereby optimizing metabolic flexibility and glucose utilization pathways (31-33). Meanwhile, individuals with borderline overweight or substantial visceral adiposity might achieve better outcomes through body recomposition strategies-rather than solely pursuing weight reduction-by integrating specialized physical activities, balanced macronutrient intake, and routine body composition monitoring (34-39). A multidisciplinary approach that considers muscle mass, metabolic markers, and demographic factors ensures interventions are precisely tailored to individual needs, ultimately refining obesity management and reducing long-term complications (40).

Strengths and limitations

Our study’s strengths include its large sample size of 3,255 participants, which ensures robust statistical power, and the use of advanced body composition analysis to comprehensively evaluate both anthropometric and compositional parameters. Additionally, the inclusion of both sexes allowed for an exploration of sex-specific differences, enhancing the relevance of the findings. However, the study’s cross-sectional nature and single-center design may limit the generalizability of results to other populations. The predominance of female participants, while reflective of typical clinical populations, may also affect the broader applicability of our findings to male populations. Future research should focus on longitudinal studies to evaluate how these relationships evolve over time and with intervention, particularly focusing on the role of PhA in predicting outcomes in weight management programs.

Conclusions

In conclusion, our study in a Colombian population reveals important sex-specific differences in body composition patterns and demonstrates that while BMI serves as an adequate predictor of adiposity, it falls short in assessing muscular mass. Traditional weight- classification systems show limitations in accurately categorizing patients, as evidenced by the distinct behavior of fat and muscle compartments in our linear regression analyses. Notably, there are significant differences between men and women that cannot be adequately captured using traditional anthropometric measurements, further highlighting the need for more precise evaluation methods. The emergence of PhA as a robust predictor of muscle mass, but not fat mass, underscores its potential utility in clinical assessment. These findings emphasize the critical importance of incorporating comprehensive body composition analysis in clinical practice, moving beyond simple weight-based metrics. Future studies should focus on integrating detailed body composition assessment into routine clinical care, particularly in diverse populations where standard anthropometric assumptions may not apply, to better inform personalized treatment strategies and improve patient outcomes.

Authors’ contributions

Ricardo Rosero-Revelo (Conceptualization, Study design, Writing - original draft, Supervision, review & editing), Mateo Tamayo (Data analysis, Data interpretation, Methodology, Writing), Marcio L. Griebeler (Original draft, Supervision, review & editing).

Ethical considerations

This study was conducted in strict accordance with international ethical standards. Given its design as an observational study involving no direct contact with participants and the exclusive use of de-identified data, ethical review was not required as per institutional guidelines. Nonetheless, all procedures were carried out with a commitment to maintain data confidentiality and participant anonymity, ensuring compliance with ethical standards.

Funding

This study did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. The resources utilized for conducting this study were solely from the internal funds of the participating institutes.

Conflicts of interest

The authors declare that there are no conflicts of interest regarding the publication of this paper. All participants have contributed impartially, free from financial or personal influences that could undermine the integrity of this work.

Acknowledgements

The authors would like to extend their appreciation to the Colombian Association of Endocrinology, Diabetes, and Metabolism for their invaluable support in presenting the research findings at Obesity Week 2024, which significantly contributed to the dissemination and discussion of the work detailed in this article.