Results

Participant characteristics

A total of 120 adolescents with overweight and obesity were included, comprising 59 (49.2%) females and 61 (50.8%) males. The median age was 13.0 years (IQR 10.8-15.0 years), with females being significantly older than males (13.1 (11.2-16.4) vs. 12.0 (10.0-14.9) years; p = 0.029). The distribution of obesity categories was as follows: overweight, 41 (34.2%); and obese, 79 (65.8%), with no significant sex-based difference (p = 0.22). The median BMI was 27.1 kg/m² (IQR 23.7-30.9 kg/m²) and did not differ significantly between sexes (p = 0.426). Body composition analysis using BIA demonstrated significant sex-based differences in BF%, SCF, SMM, and fat mass index (FMI), with females exhibiting higher adiposity measures. However, VF% did not differ significantly between females and males (12.0 (10.0-17.0) vs. 12.0 (7.0-16.0); p = 0.274). Detailed baseline characteristics are presented in Table 1.

Variable | Female n = 59 (49.2 %) | Male n = 61 (50.8 %) | Total n = 120 (100%) | Statistic | p-value
Age (years) | 13.1 (11.2 to 16.4) | 12.0 (10.0 to 14.9) | 13.0 (10.8 to 15.0) | 1385 | 0.029
Weight (kg) | 61.4 (49.9 to 72.5) | 62.0 (46.4 to 76.0) | 61.7 (48.0 to 75.2) | 1764 | 0.852
Height (cm) | 149.7 (141.9 to 155.5) | 151.0 (138.2 to 162.0) | 150.0 (140.0 to 160.4) | 1664 | 0.477
BMI (kg/m2) | 27.8 (23.6 to 31.8) | 26.8 (23.7 to 30.4) | 27.1 (23.7 to 30.9) | 1648 | 0.426
Obesity Class n (%)
Overweight | 17 (28.8) | 24 (39.3) | 41 (34.2) | 1.51 | 0.22
Obesity | 42 (71.2) | 37 (60.7) | 79 (65.8)
WC (cm) | 90.0 (80.0 to 97.5) | 86.0 (79.0 to 98.0) | 88.0 (79.0 to 98.0) | 1624 | 0.355
Hip Circumference (cm) | 97.0 (84.0 to 110.0) | 90.0 (84.0 to 105.0) | 94.0 (84.0 to 108.2) | 1571 | 0.231
WHR | 0.9 (0.9 to 1.0) | 0.9 (0.9 to 1.0) | 0.9 (0.9 to 1.0) | 1683 | 0.538
WHtR | 0.6 (0.6 to 0.6) | 0.6 (0.5 to 0.6) | 0.6 (0.5 to 0.6) | 1514 | 0.134
Conicity Index | 1.3 (1.2 to 1.3) | 1.3 (1.2 to 1.3) | 1.3 (1.2 to 1.3) | 1775 | 0.898
AVI | 21.7 (17.2 to 25.5) | 19.8 (16.8 to 25.8) | 20.8 (16.8 to 25.8) | 1631 | 0.375
Body Fat (%) | 44.0 (38.0 to 48.9) | 36.5 (28.6 to 41.0) | 40.0 (33.8 to 45.3) | 914 | <0.001
SCF | 42.7 (26.9 to 48.6) | 26.0 (20.0 to 39.0) | 34.5 (22.9 to 44.0) | 1009 | <0.001
SMM | 49.0 (30.6 to 56.0) | 30.6 (27.9 to 42.8) | 37.0 (29.0 to 52.0) | 956 | <0.001
Visceral Fat (%) | 12.0 (10.0 to 17.0) | 12.0 (7.0 to 16.0) | 12.0 (10.0 to 17.0) | 1592 | 0.274
Fat Mass (kg) | 26.3 (18.3 to 34.3) | 20.8 (15.7 to 30.6) | 22.8 (16.0 to 32.7) | 1346 | 0.017
Fat Free Mass (kg) | 35.5 (29.0 to 41.1) | 40.3 (29.6 to 46.1) | 36.8 (29.2 to 44.0) | 1419 | 0.046
FMI | 12.3 (9.2 to 14.8) | 9.0 (7.5 to 12.3) | 10.5 (8.1 to 13.8) | 1149 | 0.001
SMR Stage
0 | 7 (11.9) | 19 (31.1) | 26 (21.7) | 11.46 | 0.012
1 | 16 (27.1) | 20 (32.8) | 36 (30.0)
2 | 11 (18.6) | 11 (18.0) | 22 (18.3)
3 | 15 (25.4) | 9 (14.8) | 24 (20.0)
4 | 10 (16.9) | 2 (3.3) | 12 (10.0)
Systolic BP (mmHg) | 122.0 (114.0 to 127.0) | 118.0 (110.0 to 128.0) | 120.0 (112.0 to 128.0) | 1695 | 0.58
Diastolic BP(mmHg) | 80.0 (70.0 to 86.0) | 80.0 (70.0 to 84.0) | 80.0 (70.0 to 84.0) | 1666 | 0.477

Correlation between anthropometric indices and body composition

Spearman’s rank correlation analysis demonstrated significant positive associations between most anthropometric indices and VF%. WC and AVI showed the comparatively highest correlations with VF% among the indices studied (both ρ = 0.707, p < 0.001), a moderate-to-strong association, followed by BMI (ρ = 0.696, p < 0.001). Among derived indices, WHtR showed a moderate correlation (ρ = 0.417, p < 0.001), while WHR demonstrated a weaker but significant association (ρ = 0.387, p < 0.001). In contrast, the CI showed the weakest association with VF% among the indices studied (ρ = 0.267, p = 0.003). The CI did not correlate significantly with BF% (ρ = 0.094, p = 0.306) or FMI (ρ = 0.002, p = 0.985), suggesting limited ability to reflect overall adiposity. In contrast, FMI correlated strongly with BMI (ρ = 0.775, p < 0.001) and BF% (ρ = 0.922, p < 0.001). BF% also correlated significantly with BMI, WC, and WHR. The correlation matrix is presented in Figure 1.

Figure 1: Heatmap of Spearman rank correlation coefficientsCorrelation coefficient values are Spearman's rho.BMI: body mass index; WC: waist circumference; WHR: waist-to-hip ratio; WHtR: waist-to-height ratio; AVI: abdominal volume index; SCF: subcutaneous fat; SMM: skeletal muscle mass; VF%: visceral fat percentage; FMI: fat mass index; BAI: body adiposity index; CI: conicity index

Multiple linear regression analysis

Multiple linear regression models were developed to evaluate predictors of VF%. The base model, including age, sex, and BMI, explained 53.4% of the variance in visceral fat (R² = 0.534, adjusted R² = 0.521). The addition of WC significantly improved model performance (ΔR² = 0.091, p < 0.001), yielding a final R² of 0.624 and improved model fit. The AVI provided a comparable improvement (ΔR² = 0.087, p < 0.001), consistent with its strong mathematical dependence on WC. In comparison, the addition of the CI produced a smaller but statistically significant improvement relative to the base model (ΔR² = 0.049, p < 0.001). Other indices, including WHR, WHtR, and BAI, did not significantly enhance predictive performance. The most parsimonious clinically relevant model included age, sex, BMI, and WC. Model comparisons are presented in Table 2.

Model | R | R² | Adjusted R² | AIC | RMSE | F | df1 | df2 | p value | ΔR² | p value (ΔR²)
Base (Age+Sex+BMI) | 0.73 | 0.53 | 0.521 | 678 | 3.92 | 44 | 3 | 116 | < .001 | - | -
Base+WC | 0.79 | 0.62 | 0.611 | 655 | 3.52 | 48 | 4 | 115 | < .001 | 0.091 | < .001
Base+WHR | 0.73 | 0.54 | 0.519 | 680 | 3.91 | 33 | 4 | 115 | < .001 | 0.002 | 0.529
Base+WHtR | 0.74 | 0.54 | 0.527 | 678 | 3.88 | 34 | 4 | 115 | < .001 | 0.009 | 0.134
Base+CI | 0.76 | 0.58 | 0.568 | 667 | 3.71 | 40 | 4 | 115 | < .001 | 0.049 | < .001
Base+BAI | 0.73 | 0.54 | 0.519 | 680 | 3.91 | 33 | 4 | 115 | < .001 | 0.002 | 0.534
Base+AVI | 0.79 | 0.62 | 0.607 | 656 | 3.54 | 47 | 4 | 115 | < .001 | 0.087 | < .001
Base+WC+CI | 0.8 | 0.63 | 0.615 | 654 | 3.48 | 39 | 5 | 114 | < .001 | 0.098 | < .001
Base+WC+AVI | 0.79 | 0.63 | 0.61 | 656 | 3.51 | 38 | 5 | 114 | < .001 | 0.093 | < .001

Regression coefficients of the final model

In the final model including age, sex, BMI, and WC, BMI was the strongest independent predictor of VF%, followed by waist circumference. The unstandardized coefficient for BMI was 0.499 (standardized β = 0.454, p < 0.001), and the unstandardized coefficient for WC was 0.188 (standardized β = 0.391, p < 0.001). Thus, each unit increase in BMI was associated with a 0.499-unit increase in VF%, and each centimeter increase in WC was associated with a 0.188-unit increase, independent of the other variables. Age and sex were not statistically significant predictors but were retained in the model due to biological relevance. The derived prediction equation was:

[ \text{Visceral Fat (%)} = -19.879 + 0.499 \times BMI + 0.188 \times WC + 0.172 \times Age - 0.253 \times \text{Sex}_{\text{Male}} ]

Detailed regression coefficients are presented in Table 3.

Predictor | Estimate | SE | Lower 95% CI | Upper 95% CI | t value | p value | Stand. Estimate
Intercept | -19.879 | 2.6 | -25.029 | -14.728 | -7.65 | < .001 | -
Age | 0.172 | 0.156 | -0.136 | 0.48 | 1.106 | 0.271 | 0.079
Sex: Male | -0.253 | 0.674 | -1.588 | 1.083 | -0.38 | 0.709 | -0.043
BMI | 0.499 | 0.076 | 0.348 | 0.649 | 6.577 | < .001 | 0.454
WC (cm) | 0.188 | 0.036 | 0.117 | 0.259 | 5.264 | < .001 | 0.391