Threshold robustness of bioimpedance-derived adiposity classification across BMI categories in Chinese adults: a cross-sectional study

Background Body mass index (BMI) and bioelectrical impedance analysis (BIA)-derived adiposity indicators are frequently combined in clinical nutrition assessment. However, clinical classification based on device-generated thresholds may vary depending on the definition applied. This study evaluated the threshold robustness of BMI-BIA adiposity classification. Methods This secondary cross-sectional analysis included 93 adults with baseline body-composition measurements recorded during 2024. Participants were described using Chinese BMI categories (normal weight, overweight, and obesity), and threshold-robustness analyses were conducted in BMI < 24 kg/m 2 and BMI ≥ 24 kg/m 2 strata. Percent body fat (PBF) and visceral fat area (VFA) classifications based on device-reported thresholds were compared with alternative thresholds. A three-state framework distinguished stable non-elevated, threshold-contingent elevated, and threshold-robust elevated classifications. Results Threshold sensitivity differed between adiposity indicators. Under the specific threshold pairs evaluated, PBF showed greater threshold-dependent reclassification than VFA. Among participants with BMI < 24 kg/m 2 , elevated PBF decreased from 25.9% under device-report thresholds to 3.7% under the alternative threshold, corresponding to 22.2 percentage points of threshold-contingent classification. Among participants with BMI ≥ 24 kg/m 2 , elevated PBF decreased from 100 to 60.6%, resulting in 39.4% threshold-contingent classification. VFA classification was more stable under the evaluated threshold pair, with reclassification concentrated among higher-BMI women. In supplementary standardized threshold-shift analyses, the between-indicator difference was attenuated when threshold displacement was normalized by indicator variability. Conclusion BMI-BIA adiposity classification was influenced by threshold selection. Under the evaluated threshold definitions, PBF showed greater threshold-dependent reclassification than VFA; however, this finding reflects the specific threshold scenarios examined and should not be interpreted as an intrinsic superiority of one indicator over another. The threshold robustness framework may help distinguish stable from threshold-dependent classifications and requires further validation in larger cohorts.