Assessment of a proposed BMI formula in predicting body fat percentage among Filipino young adults

Obesity is a growing global health concern, particularly in Asia, including the Philippines. Body Mass Index (BMI), calculated using the Quetelet index (weight in kg/(height in m)²), is a widely used anthropometric measure for assessing adiposity. However, BMI has limitations as it does not differentiate between fat and lean mass, leading to misclassification of weight status, especially in athletic individuals, or those with varying height and body proportions. Professor L. N. Trefethen proposed a modified BMI formula (BMI = 1.3(weight in kg)/(height in m)^2.5) to address this, suggesting it better approximates body size and shape by adjusting for the disproportionate impact of height in the traditional formula on shorter and taller individuals. This study aimed to compare the performance of Trefethen's modified BMI formula (BMI_T) against the traditional Keys' BMI formula (BMI_K) in predicting body fat percentage (%BF) and diagnosing overweight/obesity in a sample of Filipino young adults. The increasing prevalence of obesity in the Philippines, coupled with the limitations of the traditional BMI, creates a need for improved assessment methods. A more accurate prediction of %BF could lead to better targeted interventions and improved public health strategies.

The existing literature highlights the limitations of the traditional BMI in accurately reflecting body fat percentage. Studies have shown BMI's inability to distinguish between fat and lean mass, leading to misclassification particularly in athletic individuals and those with varying heights. Research indicates that BMI underestimates adiposity in shorter individuals and overestimates it in taller individuals. The inconsistencies in the BMI-body fat relationship across different ethnic groups and age groups further underscore the need for improved assessment methods. Several alternative anthropometric measures, like waist circumference, have been proposed, but BMI's ease of measurement and widespread use remain its primary advantages. Trefethen's modified formula offers a potential solution by adjusting the height component of the BMI calculation to account for variations in body proportions. However, epidemiological evidence supporting the use of the proposed exponent on height is lacking. The study design builds upon these existing research gaps to evaluate the utility of Trefethen's modified formula within the context of a specific population.

This cross-sectional observational study recruited 190 Filipino medical students (74 males, 116 females) aged 18-35 years. Participants were excluded if they had chronic illnesses, recent cardiovascular events, pregnancy, corticosteroid use, postural conditions, or participated in regular exercise programs. Data collection involved a questionnaire (demographic information, smoking history, alcohol intake) and direct measurements (height, weight, waist circumference, %BF using bioelectric impedance analysis). BMI was calculated using both the traditional (BMI_K) and Trefethen's modified (BMI_T) formulas. Weight classifications were determined using Asian-Pacific cutoff points. Obesity was defined as ≥25% BF in males and ≥35% BF in females. Statistical analyses included descriptive statistics, t-tests/Mann-Whitney U tests for group comparisons, Pearson's/Spearman's correlations, Cohen's kappa for agreement, robust polynomial regression (linear and quadratic models) to predict %BF using BMI (both formulas), and ROC curve analysis to assess diagnostic accuracy for overweight/obesity. Model adequacy was assessed through residual analysis, LOWESS curves, and information criteria (AIC, BIC). Post-hoc power analysis was conducted to assess the adequacy of the sample size.

The study found high correlations between both BMI types (BMI_K and BMI_T) and %BF. Correlations tended to be slightly higher in females, although this difference was not significant. Agreement between weight classifications based on the two BMI formulas was high, particularly in males. Robust polynomial regression revealed that quadratic models provided a better fit than linear models for both sexes in predicting %BF, indicated by higher adjusted R² and lower AIC and BIC values. Both BMI_K and BMI_T were significant predictors of %BF, with waist circumference also being significant for females. Regarding diagnostic accuracy for overweight/obesity, using a BMI cutoff of 23.0, both BMI formulas showed high sensitivity and negative predictive value, but poor specificity, particularly in males. Optimal cutoff values for both BMIs were identified using ROC analysis, yielding high AUROCs (>0.90) for both sexes, although the difference between BMI_K and BMI_T AUROCs was not statistically significant. Post-hoc power analysis indicated sufficient sample size for the primary objective (predicting %BF) but underpowered the comparison of AUROCs between BMI_K and BMI_T.

This study demonstrated that both the traditional and Trefethen's modified BMI formulas are significant predictors of body fat percentage and can adequately discriminate between normal and overweight/obese states in young Filipino adults. The quadratic relationship between BMI and %BF supports findings in other Asian populations. While both formulas showed good predictive ability, the relatively poor specificity highlights the inherent limitations of BMI in differentiating fat and lean mass. The observed changes in weight classification between BMI_K and BMI_T, especially in individuals below the average height used in Trefethen's formula, warrant further investigation into potential long-term health implications. The study's findings underscore the complexity of BMI-body fat relationships, particularly in diverse populations. Further research is needed to determine if the optimal BMI cutoff values and formulas vary across different ethnic groups and age ranges.

This study provides valuable insights into the performance of the traditional and modified BMI formulas for predicting body fat percentage and diagnosing overweight/obesity in Filipino young adults. While both formulas showed promise, the limitations of BMI in distinguishing fat and lean mass are reaffirmed. Future prospective studies should investigate the long-term health implications of using different BMI formulas and explore the development of population-specific BMI prediction models.

The study's limitations include the use of a non-probability sample (self-selection bias), which may overrepresent overweight/obese individuals. The small sample size, while adequate for the primary objective, underpowered the comparison of AUROC values between BMI formulas. The study's findings are not generalizable to older adults, physically active individuals, or other ethnic groups. The use of bioelectrical impedance analysis for %BF measurement should also be considered, while acknowledging its limitations compared to the gold standard of DXA. Finally, the exclusion of individuals with pre-existing conditions prevents an assessment of the effect of comorbidities on the BMI-%BF relationship.