Error reduction as a calibration strategy for body composition measurements: A comparison between bioelectrical...

Accurate diagnosis of obesity relies on valid body composition assessment methods. Dual-energy X-ray absorptiometry (DEXA) is the clinical gold standard for assessing body composition. Still, it is costly, leading to the widespread use of bioelectrical impedance analysis (BIA) scales, which often yield values that differ from those obtained with DEXA. This study aimed to: (a) quantify the error between anthropometric variables measured by BIA and DEXA; (b) identify variables with the highest and lowest errors, including sex- and body mass index (BMI)- specific analyses; and (c) propose a percentile-based calibration and ordinary least squares (OLS) linear regression method to reduce measurement error.

Sixty-one participants (33 men, 28 women; aged 24-63 years) underwent height, weight, BMI, and body composition assessments using BIA and DEXA. A mirrored dataset was created using both methods. A linear-percentile calibration and OLS approach was used to align empirical quantiles, adjusting BIA measurements to DEXA standards.

After calibration, the variables showed the highest predictive accuracy, while the trunk and limb BFM variables performed the worst, with trunk FFM (fat-free mass) improving only by the OLS method. Calibration reduced errors in most BMI and sex groups, except for trunk FFM and leg FFM, for which calibration showed no benefit.

BIA demonstrates systematic measurement errors compared to DEXA. However, calibration through percentile alignment led to notable improvements, with the regression models showing a significant improvement in agreement between methods within this sample. offering a practical strategy to enhance the accuracy of BIA-based assessments in clinical and research contexts.